1 # Ocean Interpreter - Jamison Creek version
3 Ocean is intended to be a compiled language, so this interpreter is
4 not targeted at being the final product. It is, rather, an intermediate
5 stage and fills that role in two distinct ways.
7 Firstly, it exists as a platform to experiment with the early language
8 design. An interpreter is easy to write and easy to get working, so
9 the barrier for entry is lower if I aim to start with an interpreter.
11 Secondly, the plan for the Ocean compiler is to write it in the
12 [Ocean language](http://ocean-lang.org). To achieve this we naturally
13 need some sort of boot-strap process and this interpreter - written in
14 portable C - will fill that role. It will be used to bootstrap the
17 Two features that are not needed to fill either of these roles are
18 performance and completeness. The interpreter only needs to be fast
19 enough to run small test programs and occasionally to run the compiler
20 on itself. It only needs to be complete enough to test aspects of the
21 design which are developed before the compiler is working, and to run
22 the compiler on itself. Any features not used by the compiler when
23 compiling itself are superfluous. They may be included anyway, but
26 Nonetheless, the interpreter should end up being reasonably complete,
27 and any performance bottlenecks which appear and are easily fixed, will
32 This third version of the interpreter exists to test out some initial
33 ideas relating to types. Particularly it adds arrays (indexed from
34 zero) and simple structures. Basic control flow and variable scoping
35 are already fairly well established, as are basic numerical and
38 Some operators that have only recently been added, and so have not
39 generated all that much experience yet are "and then" and "or else" as
40 short-circuit Boolean operators, and the "if ... else" trinary
41 operator which can select between two expressions based on a third
42 (which appears syntactically in the middle).
44 The "func" clause currently only allows a "main" function to be
45 declared. That will be extended when proper function support is added.
47 An element that is present purely to make a usable language, and
48 without any expectation that they will remain, is the "print" statement
49 which performs simple output.
51 The current scalar types are "number", "Boolean", and "string".
52 Boolean will likely stay in its current form, the other two might, but
53 could just as easily be changed.
57 Versions of the interpreter which obviously do not support a complete
58 language will be named after creeks and streams. This one is Jamison
61 Once we have something reasonably resembling a complete language, the
62 names of rivers will be used.
63 Early versions of the compiler will be named after seas. Major
64 releases of the compiler will be named after oceans. Hopefully I will
65 be finished once I get to the Pacific Ocean release.
69 As well as parsing and executing a program, the interpreter can print
70 out the program from the parsed internal structure. This is useful
71 for validating the parsing.
72 So the main requirements of the interpreter are:
74 - Parse the program, possibly with tracing,
75 - Analyse the parsed program to ensure consistency,
77 - Execute the "main" function in the program, if no parsing or
78 consistency errors were found.
80 This is all performed by a single C program extracted with
83 There will be two formats for printing the program: a default and one
84 that uses bracketing. So a `--bracket` command line option is needed
85 for that. Normally the first code section found is used, however an
86 alternate section can be requested so that a file (such as this one)
87 can contain multiple programs. This is effected with the `--section`
90 This code must be compiled with `-fplan9-extensions` so that anonymous
91 structures can be used.
93 ###### File: oceani.mk
95 myCFLAGS := -Wall -g -fplan9-extensions
96 CFLAGS := $(filter-out $(myCFLAGS),$(CFLAGS)) $(myCFLAGS)
97 myLDLIBS:= libparser.o libscanner.o libmdcode.o -licuuc
98 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
100 all :: $(LDLIBS) oceani
101 oceani.c oceani.h : oceani.mdc parsergen
102 ./parsergen -o oceani --LALR --tag Parser oceani.mdc
103 oceani.mk: oceani.mdc md2c
106 oceani: oceani.o $(LDLIBS)
107 $(CC) $(CFLAGS) -o oceani oceani.o $(LDLIBS)
109 ###### Parser: header
111 struct parse_context;
113 struct parse_context {
114 struct token_config config;
122 #define container_of(ptr, type, member) ({ \
123 const typeof( ((type *)0)->member ) *__mptr = (ptr); \
124 (type *)( (char *)__mptr - offsetof(type,member) );})
126 #define config2context(_conf) container_of(_conf, struct parse_context, \
129 ###### Parser: reduce
130 struct parse_context *c = config2context(config);
138 #include <sys/mman.h>
157 static char Usage[] =
158 "Usage: oceani --trace --print --noexec --brackets --section=SectionName prog.ocn\n";
159 static const struct option long_options[] = {
160 {"trace", 0, NULL, 't'},
161 {"print", 0, NULL, 'p'},
162 {"noexec", 0, NULL, 'n'},
163 {"brackets", 0, NULL, 'b'},
164 {"section", 1, NULL, 's'},
167 const char *options = "tpnbs";
169 static void pr_err(char *msg) // NOTEST
171 fprintf(stderr, "%s\n", msg); // NOTEST
174 int main(int argc, char *argv[])
179 struct section *s = NULL, *ss;
180 char *section = NULL;
181 struct parse_context context = {
183 .ignored = (1 << TK_mark),
184 .number_chars = ".,_+- ",
189 int doprint=0, dotrace=0, doexec=1, brackets=0;
191 while ((opt = getopt_long(argc, argv, options, long_options, NULL))
194 case 't': dotrace=1; break;
195 case 'p': doprint=1; break;
196 case 'n': doexec=0; break;
197 case 'b': brackets=1; break;
198 case 's': section = optarg; break;
199 default: fprintf(stderr, Usage);
203 if (optind >= argc) {
204 fprintf(stderr, "oceani: no input file given\n");
207 fd = open(argv[optind], O_RDONLY);
209 fprintf(stderr, "oceani: cannot open %s\n", argv[optind]);
212 context.file_name = argv[optind];
213 len = lseek(fd, 0, 2);
214 file = mmap(NULL, len, PROT_READ, MAP_SHARED, fd, 0);
215 s = code_extract(file, file+len, pr_err);
217 fprintf(stderr, "oceani: could not find any code in %s\n",
222 ## context initialization
225 for (ss = s; ss; ss = ss->next) {
226 struct text sec = ss->section;
227 if (sec.len == strlen(section) &&
228 strncmp(sec.txt, section, sec.len) == 0)
232 fprintf(stderr, "oceani: cannot find section %s\n",
239 fprintf(stderr, "oceani: no code found in requested section\n"); // NOTEST
240 goto cleanup; // NOTEST
243 parse_oceani(ss->code, &context.config, dotrace ? stderr : NULL);
245 resolve_consts(&context);
246 prepare_types(&context);
247 if (!context.parse_error && !analyse_funcs(&context)) {
248 fprintf(stderr, "oceani: type error in program - not running.\n");
249 context.parse_error = 1;
257 if (doexec && !context.parse_error)
258 interp_main(&context, argc - optind, argv + optind);
261 struct section *t = s->next;
266 // FIXME parser should pop scope even on error
267 while (context.scope_depth > 0)
271 ## free context types
272 ## free context storage
273 exit(context.parse_error ? 1 : 0);
278 The four requirements of parse, analyse, print, interpret apply to
279 each language element individually so that is how most of the code
282 Three of the four are fairly self explanatory. The one that requires
283 a little explanation is the analysis step.
285 The current language design does not require the types of variables to
286 be declared, but they must still have a single type. Different
287 operations impose different requirements on the variables, for example
288 addition requires both arguments to be numeric, and assignment
289 requires the variable on the left to have the same type as the
290 expression on the right.
292 Analysis involves propagating these type requirements around and
293 consequently setting the type of each variable. If any requirements
294 are violated (e.g. a string is compared with a number) or if a
295 variable needs to have two different types, then an error is raised
296 and the program will not run.
298 If the same variable is declared in both branchs of an 'if/else', or
299 in all cases of a 'switch' then the multiple instances may be merged
300 into just one variable if the variable is referenced after the
301 conditional statement. When this happens, the types must naturally be
302 consistent across all the branches. When the variable is not used
303 outside the if, the variables in the different branches are distinct
304 and can be of different types.
306 Undeclared names may only appear in "use" statements and "case" expressions.
307 These names are given a type of "label" and a unique value.
308 This allows them to fill the role of a name in an enumerated type, which
309 is useful for testing the `switch` statement.
311 As we will see, the condition part of a `while` statement can return
312 either a Boolean or some other type. This requires that the expected
313 type that gets passed around comprises a type and a flag to indicate
314 that `Tbool` is also permitted.
316 As there are, as yet, no distinct types that are compatible, there
317 isn't much subtlety in the analysis. When we have distinct number
318 types, this will become more interesting.
322 When analysis discovers an inconsistency it needs to report an error;
323 just refusing to run the code ensures that the error doesn't cascade,
324 but by itself it isn't very useful. A clear understanding of the sort
325 of error message that are useful will help guide the process of
328 At a simplistic level, the only sort of error that type analysis can
329 report is that the type of some construct doesn't match a contextual
330 requirement. For example, in `4 + "hello"` the addition provides a
331 contextual requirement for numbers, but `"hello"` is not a number. In
332 this particular example no further information is needed as the types
333 are obvious from local information. When a variable is involved that
334 isn't the case. It may be helpful to explain why the variable has a
335 particular type, by indicating the location where the type was set,
336 whether by declaration or usage.
338 Using a recursive-descent analysis we can easily detect a problem at
339 multiple locations. In "`hello:= "there"; 4 + hello`" the addition
340 will detect that one argument is not a number and the usage of `hello`
341 will detect that a number was wanted, but not provided. In this
342 (early) version of the language, we will generate error reports at
343 multiple locations, so the use of `hello` will report an error and
344 explain were the value was set, and the addition will report an error
345 and say why numbers are needed. To be able to report locations for
346 errors, each language element will need to record a file location
347 (line and column) and each variable will need to record the language
348 element where its type was set. For now we will assume that each line
349 of an error message indicates one location in the file, and up to 2
350 types. So we provide a `printf`-like function which takes a format, a
351 location (a `struct exec` which has not yet been introduced), and 2
352 types. "`%1`" reports the first type, "`%2`" reports the second. We
353 will need a function to print the location, once we know how that is
354 stored. e As will be explained later, there are sometimes extra rules for
355 type matching and they might affect error messages, we need to pass those
358 As well as type errors, we sometimes need to report problems with
359 tokens, which might be unexpected or might name a type that has not
360 been defined. For these we have `tok_err()` which reports an error
361 with a given token. Each of the error functions sets the flag in the
362 context so indicate that parsing failed.
366 static void fput_loc(struct exec *loc, FILE *f);
367 static void type_err(struct parse_context *c,
368 char *fmt, struct exec *loc,
369 struct type *t1, int rules, struct type *t2);
371 ###### core functions
373 static void type_err(struct parse_context *c,
374 char *fmt, struct exec *loc,
375 struct type *t1, int rules, struct type *t2)
377 fprintf(stderr, "%s:", c->file_name);
378 fput_loc(loc, stderr);
379 for (; *fmt ; fmt++) {
386 case '%': fputc(*fmt, stderr); break; // NOTEST
387 default: fputc('?', stderr); break; // NOTEST
389 type_print(t1, stderr);
392 type_print(t2, stderr);
401 static void tok_err(struct parse_context *c, char *fmt, struct token *t)
403 fprintf(stderr, "%s:%d:%d: %s: %.*s\n", c->file_name, t->line, t->col, fmt,
404 t->txt.len, t->txt.txt);
408 ## Entities: declared and predeclared.
410 There are various "things" that the language and/or the interpreter
411 needs to know about to parse and execute a program. These include
412 types, variables, values, and executable code. These are all lumped
413 together under the term "entities" (calling them "objects" would be
414 confusing) and introduced here. The following section will present the
415 different specific code elements which comprise or manipulate these
420 Executables can be lots of different things. In many cases an
421 executable is just an operation combined with one or two other
422 executables. This allows for expressions and lists etc. Other times an
423 executable is something quite specific like a constant or variable name.
424 So we define a `struct exec` to be a general executable with a type, and
425 a `struct binode` which is a subclass of `exec`, forms a node in a
426 binary tree, and holds an operation. There will be other subclasses,
427 and to access these we need to be able to `cast` the `exec` into the
428 various other types. The first field in any `struct exec` is the type
429 from the `exec_types` enum.
432 #define cast(structname, pointer) ({ \
433 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
434 if (__mptr && *__mptr != X##structname) abort(); \
435 (struct structname *)( (char *)__mptr);})
437 #define new(structname) ({ \
438 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
439 __ptr->type = X##structname; \
440 __ptr->line = -1; __ptr->column = -1; \
443 #define new_pos(structname, token) ({ \
444 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
445 __ptr->type = X##structname; \
446 __ptr->line = token.line; __ptr->column = token.col; \
455 enum exec_types type;
464 struct exec *left, *right;
469 static int __fput_loc(struct exec *loc, FILE *f)
473 if (loc->line >= 0) {
474 fprintf(f, "%d:%d: ", loc->line, loc->column);
477 if (loc->type == Xbinode)
478 return __fput_loc(cast(binode,loc)->left, f) ||
479 __fput_loc(cast(binode,loc)->right, f); // NOTEST
482 static void fput_loc(struct exec *loc, FILE *f)
484 if (!__fput_loc(loc, f))
485 fprintf(f, "??:??: ");
488 Each different type of `exec` node needs a number of functions defined,
489 a bit like methods. We must be able to free it, print it, analyse it
490 and execute it. Once we have specific `exec` types we will need to
491 parse them too. Let's take this a bit more slowly.
495 The parser generator requires a `free_foo` function for each struct
496 that stores attributes and they will often be `exec`s and subtypes
497 there-of. So we need `free_exec` which can handle all the subtypes,
498 and we need `free_binode`.
502 static void free_binode(struct binode *b)
511 ###### core functions
512 static void free_exec(struct exec *e)
523 static void free_exec(struct exec *e);
525 ###### free exec cases
526 case Xbinode: free_binode(cast(binode, e)); break;
530 Printing an `exec` requires that we know the current indent level for
531 printing line-oriented components. As will become clear later, we
532 also want to know what sort of bracketing to use.
536 static void do_indent(int i, char *str)
543 ###### core functions
544 static void print_binode(struct binode *b, int indent, int bracket)
548 ## print binode cases
552 static void print_exec(struct exec *e, int indent, int bracket)
558 print_binode(cast(binode, e), indent, bracket); break;
563 do_indent(indent, "/* FREE");
564 for (v = e->to_free; v; v = v->next_free) {
565 printf(" %.*s", v->name->name.len, v->name->name.txt);
566 printf("[%d,%d]", v->scope_start, v->scope_end);
567 if (v->frame_pos >= 0)
568 printf("(%d+%d)", v->frame_pos,
569 v->type ? v->type->size:0);
577 static void print_exec(struct exec *e, int indent, int bracket);
581 As discussed, analysis involves propagating type requirements around the
582 program and looking for errors.
584 So `propagate_types` is passed an expected type (being a `struct type`
585 pointer together with some `val_rules` flags) that the `exec` is
586 expected to return, and returns the type that it does return, either of
587 which can be `NULL` signifying "unknown". A `prop_err` flag set is
588 passed by reference. It has `Efail` set when an error is found, and
589 `Eretry` when the type for some element is set via propagation. If it
590 remains unchanged at `0`, then no more propagation is needed.
594 enum val_rules {Rnolabel = 1<<0, Rboolok = 1<<1, Rnoconstant = 1<<2};
595 enum prop_err {Efail = 1<<0, Eretry = 1<<1};
599 if (rules & Rnolabel)
600 fputs(" (labels not permitted)", stderr);
604 static struct type *propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
605 struct type *type, int rules);
606 ###### core functions
608 static struct type *__propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
609 struct type *type, int rules)
616 switch (prog->type) {
619 struct binode *b = cast(binode, prog);
621 ## propagate binode cases
625 ## propagate exec cases
630 static struct type *propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
631 struct type *type, int rules)
633 struct type *ret = __propagate_types(prog, c, perr, type, rules);
642 Interpreting an `exec` doesn't require anything but the `exec`. State
643 is stored in variables and each variable will be directly linked from
644 within the `exec` tree. The exception to this is the `main` function
645 which needs to look at command line arguments. This function will be
646 interpreted separately.
648 Each `exec` can return a value combined with a type in `struct lrval`.
649 The type may be `Tnone` but must be non-NULL. Some `exec`s will return
650 the location of a value, which can be updated, in `lval`. Others will
651 set `lval` to NULL indicating that there is a value of appropriate type
655 static struct value interp_exec(struct parse_context *c, struct exec *e,
656 struct type **typeret);
657 ###### core functions
661 struct value rval, *lval;
664 /* If dest is passed, dtype must give the expected type, and
665 * result can go there, in which case type is returned as NULL.
667 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
668 struct value *dest, struct type *dtype);
670 static struct value interp_exec(struct parse_context *c, struct exec *e,
671 struct type **typeret)
673 struct lrval ret = _interp_exec(c, e, NULL, NULL);
675 if (!ret.type) abort();
679 dup_value(ret.type, ret.lval, &ret.rval);
683 static struct value *linterp_exec(struct parse_context *c, struct exec *e,
684 struct type **typeret)
686 struct lrval ret = _interp_exec(c, e, NULL, NULL);
688 if (!ret.type) abort();
692 free_value(ret.type, &ret.rval);
696 /* dinterp_exec is used when the destination type is certain and
697 * the value has a place to go.
699 static void dinterp_exec(struct parse_context *c, struct exec *e,
700 struct value *dest, struct type *dtype,
703 struct lrval ret = _interp_exec(c, e, dest, dtype);
707 free_value(dtype, dest);
709 dup_value(dtype, ret.lval, dest);
711 memcpy(dest, &ret.rval, dtype->size);
714 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
715 struct value *dest, struct type *dtype)
717 /* If the result is copied to dest, ret.type is set to NULL */
719 struct value rv = {}, *lrv = NULL;
722 rvtype = ret.type = Tnone;
732 struct binode *b = cast(binode, e);
733 struct value left, right, *lleft;
734 struct type *ltype, *rtype;
735 ltype = rtype = Tnone;
737 ## interp binode cases
739 free_value(ltype, &left);
740 free_value(rtype, &right);
750 ## interp exec cleanup
756 Values come in a wide range of types, with more likely to be added.
757 Each type needs to be able to print its own values (for convenience at
758 least) as well as to compare two values, at least for equality and
759 possibly for order. For now, values might need to be duplicated and
760 freed, though eventually such manipulations will be better integrated
763 Rather than requiring every numeric type to support all numeric
764 operations (add, multiply, etc), we allow types to be able to present
765 as one of a few standard types: integer, float, and fraction. The
766 existence of these conversion functions eventually enable types to
767 determine if they are compatible with other types, though such types
768 have not yet been implemented.
770 Named type are stored in a simple linked list. Objects of each type are
771 "values" which are often passed around by value.
773 There are both explicitly named types, and anonymous types. Anonymous
774 cannot be accessed by name, but are used internally and have a name
775 which might be reported in error messages.
782 ## value union fields
791 void (*init)(struct type *type, struct value *val);
792 void (*prepare_type)(struct parse_context *c, struct type *type, int parse_time);
793 void (*print)(struct type *type, struct value *val, FILE *f);
794 void (*print_type)(struct type *type, FILE *f);
795 int (*cmp_order)(struct type *t1, struct type *t2,
796 struct value *v1, struct value *v2);
797 int (*cmp_eq)(struct type *t1, struct type *t2,
798 struct value *v1, struct value *v2);
799 void (*dup)(struct type *type, struct value *vold, struct value *vnew);
800 void (*free)(struct type *type, struct value *val);
801 void (*free_type)(struct type *t);
802 long long (*to_int)(struct value *v);
803 double (*to_float)(struct value *v);
804 int (*to_mpq)(mpq_t *q, struct value *v);
813 struct type *typelist;
820 static struct type *find_type(struct parse_context *c, struct text s)
822 struct type *t = c->typelist;
824 while (t && (t->anon ||
825 text_cmp(t->name, s) != 0))
830 static struct type *_add_type(struct parse_context *c, struct text s,
831 struct type *proto, int anon)
835 n = calloc(1, sizeof(*n));
839 n->next = c->typelist;
844 static struct type *add_type(struct parse_context *c, struct text s,
847 return _add_type(c, s, proto, 0);
850 static struct type *add_anon_type(struct parse_context *c,
851 struct type *proto, char *name, ...)
857 vasprintf(&t.txt, name, ap);
859 t.len = strlen(name);
860 return _add_type(c, t, proto, 1);
863 static void free_type(struct type *t)
865 /* The type is always a reference to something in the
866 * context, so we don't need to free anything.
870 static void free_value(struct type *type, struct value *v)
874 memset(v, 0x5a, type->size);
878 static void type_print(struct type *type, FILE *f)
881 fputs("*unknown*type*", f); // NOTEST
882 else if (type->name.len && !type->anon)
883 fprintf(f, "%.*s", type->name.len, type->name.txt);
884 else if (type->print_type)
885 type->print_type(type, f);
887 fputs("*invalid*type*", f);
890 static void val_init(struct type *type, struct value *val)
892 if (type && type->init)
893 type->init(type, val);
896 static void dup_value(struct type *type,
897 struct value *vold, struct value *vnew)
899 if (type && type->dup)
900 type->dup(type, vold, vnew);
903 static int value_cmp(struct type *tl, struct type *tr,
904 struct value *left, struct value *right)
906 if (tl && tl->cmp_order)
907 return tl->cmp_order(tl, tr, left, right);
908 if (tl && tl->cmp_eq) // NOTEST
909 return tl->cmp_eq(tl, tr, left, right); // NOTEST
913 static void print_value(struct type *type, struct value *v, FILE *f)
915 if (type && type->print)
916 type->print(type, v, f);
918 fprintf(f, "*Unknown*"); // NOTEST
921 static void prepare_types(struct parse_context *c)
925 for (t = c->typelist; t; t = t->next)
927 t->prepare_type(c, t, 1);
932 static void free_value(struct type *type, struct value *v);
933 static int type_compat(struct type *require, struct type *have, int rules);
934 static void type_print(struct type *type, FILE *f);
935 static void val_init(struct type *type, struct value *v);
936 static void dup_value(struct type *type,
937 struct value *vold, struct value *vnew);
938 static int value_cmp(struct type *tl, struct type *tr,
939 struct value *left, struct value *right);
940 static void print_value(struct type *type, struct value *v, FILE *f);
942 ###### free context types
944 while (context.typelist) {
945 struct type *t = context.typelist;
947 context.typelist = t->next;
955 Type can be specified for local variables, for fields in a structure,
956 for formal parameters to functions, and possibly elsewhere. Different
957 rules may apply in different contexts. As a minimum, a named type may
958 always be used. Currently the type of a formal parameter can be
959 different from types in other contexts, so we have a separate grammar
965 Type -> IDENTIFIER ${
966 $0 = find_type(c, $1.txt);
969 "error: undefined type", &$1);
976 FormalType -> Type ${ $0 = $<1; }$
977 ## formal type grammar
981 Values of the base types can be numbers, which we represent as
982 multi-precision fractions, strings, Booleans and labels. When
983 analysing the program we also need to allow for places where no value
984 is meaningful (type `Tnone`) and where we don't know what type to
985 expect yet (type is `NULL`).
987 Values are never shared, they are always copied when used, and freed
988 when no longer needed.
990 When propagating type information around the program, we need to
991 determine if two types are compatible, where type `NULL` is compatible
992 with anything. There are two special cases with type compatibility,
993 both related to the Conditional Statement which will be described
994 later. In some cases a Boolean can be accepted as well as some other
995 primary type, and in others any type is acceptable except a label (`Vlabel`).
996 A separate function encoding these cases will simplify some code later.
998 ###### type functions
1000 int (*compat)(struct type *this, struct type *other);
1002 ###### ast functions
1004 static int type_compat(struct type *require, struct type *have, int rules)
1006 if ((rules & Rboolok) && have == Tbool)
1008 if ((rules & Rnolabel) && have == Tlabel)
1010 if (!require || !have)
1013 if (require->compat)
1014 return require->compat(require, have);
1016 return require == have;
1021 #include "parse_string.h"
1022 #include "parse_number.h"
1025 myLDLIBS := libnumber.o libstring.o -lgmp
1026 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
1028 ###### type union fields
1029 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
1031 ###### value union fields
1037 ###### ast functions
1038 static void _free_value(struct type *type, struct value *v)
1042 switch (type->vtype) {
1044 case Vstr: free(v->str.txt); break;
1045 case Vnum: mpq_clear(v->num); break;
1051 ###### value functions
1053 static void _val_init(struct type *type, struct value *val)
1055 switch(type->vtype) {
1056 case Vnone: // NOTEST
1059 mpq_init(val->num); break;
1061 val->str.txt = malloc(1);
1073 static void _dup_value(struct type *type,
1074 struct value *vold, struct value *vnew)
1076 switch (type->vtype) {
1077 case Vnone: // NOTEST
1080 vnew->label = vold->label;
1083 vnew->bool = vold->bool;
1086 mpq_init(vnew->num);
1087 mpq_set(vnew->num, vold->num);
1090 vnew->str.len = vold->str.len;
1091 vnew->str.txt = malloc(vnew->str.len);
1092 memcpy(vnew->str.txt, vold->str.txt, vnew->str.len);
1097 static int _value_cmp(struct type *tl, struct type *tr,
1098 struct value *left, struct value *right)
1102 return tl - tr; // NOTEST
1103 switch (tl->vtype) {
1104 case Vlabel: cmp = left->label == right->label ? 0 : 1; break;
1105 case Vnum: cmp = mpq_cmp(left->num, right->num); break;
1106 case Vstr: cmp = text_cmp(left->str, right->str); break;
1107 case Vbool: cmp = left->bool - right->bool; break;
1108 case Vnone: cmp = 0; // NOTEST
1113 static void _print_value(struct type *type, struct value *v, FILE *f)
1115 switch (type->vtype) {
1116 case Vnone: // NOTEST
1117 fprintf(f, "*no-value*"); break; // NOTEST
1118 case Vlabel: // NOTEST
1119 fprintf(f, "*label-%p*", v->label); break; // NOTEST
1121 fprintf(f, "%.*s", v->str.len, v->str.txt); break;
1123 fprintf(f, "%s", v->bool ? "True":"False"); break;
1128 mpf_set_q(fl, v->num);
1129 gmp_fprintf(f, "%.10Fg", fl);
1136 static void _free_value(struct type *type, struct value *v);
1138 static struct type base_prototype = {
1140 .print = _print_value,
1141 .cmp_order = _value_cmp,
1142 .cmp_eq = _value_cmp,
1144 .free = _free_value,
1147 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
1149 ###### ast functions
1150 static struct type *add_base_type(struct parse_context *c, char *n,
1151 enum vtype vt, int size)
1153 struct text txt = { n, strlen(n) };
1156 t = add_type(c, txt, &base_prototype);
1159 t->align = size > sizeof(void*) ? sizeof(void*) : size;
1160 if (t->size & (t->align - 1))
1161 t->size = (t->size | (t->align - 1)) + 1; // NOTEST
1165 ###### context initialization
1167 Tbool = add_base_type(&context, "Boolean", Vbool, sizeof(char));
1168 Tstr = add_base_type(&context, "string", Vstr, sizeof(struct text));
1169 Tnum = add_base_type(&context, "number", Vnum, sizeof(mpq_t));
1170 Tnone = add_base_type(&context, "none", Vnone, 0);
1171 Tlabel = add_base_type(&context, "label", Vlabel, sizeof(void*));
1175 We have already met values as separate objects. When manifest constants
1176 appear in the program text, that must result in an executable which has
1177 a constant value. So the `val` structure embeds a value in an
1190 ###### ast functions
1191 struct val *new_val(struct type *T, struct token tk)
1193 struct val *v = new_pos(val, tk);
1204 $0 = new_val(Tbool, $1);
1208 $0 = new_val(Tbool, $1);
1213 $0 = new_val(Tnum, $1);
1214 if (number_parse($0->val.num, tail, $1.txt) == 0)
1215 mpq_init($0->val.num); // UNTESTED
1217 tok_err(c, "error: unsupported number suffix",
1222 $0 = new_val(Tstr, $1);
1223 string_parse(&$1, '\\', &$0->val.str, tail);
1225 tok_err(c, "error: unsupported string suffix",
1230 $0 = new_val(Tstr, $1);
1231 string_parse(&$1, '\\', &$0->val.str, tail);
1233 tok_err(c, "error: unsupported string suffix",
1237 ###### print exec cases
1240 struct val *v = cast(val, e);
1241 if (v->vtype == Tstr)
1243 // FIXME how to ensure numbers have same precision.
1244 print_value(v->vtype, &v->val, stdout);
1245 if (v->vtype == Tstr)
1250 ###### propagate exec cases
1253 struct val *val = cast(val, prog);
1254 if (!type_compat(type, val->vtype, rules))
1255 type_err(c, "error: expected %1%r found %2",
1256 prog, type, rules, val->vtype);
1260 ###### interp exec cases
1262 rvtype = cast(val, e)->vtype;
1263 dup_value(rvtype, &cast(val, e)->val, &rv);
1266 ###### ast functions
1267 static void free_val(struct val *v)
1270 free_value(v->vtype, &v->val);
1274 ###### free exec cases
1275 case Xval: free_val(cast(val, e)); break;
1277 ###### ast functions
1278 // Move all nodes from 'b' to 'rv', reversing their order.
1279 // In 'b' 'left' is a list, and 'right' is the last node.
1280 // In 'rv', left' is the first node and 'right' is a list.
1281 static struct binode *reorder_bilist(struct binode *b)
1283 struct binode *rv = NULL;
1286 struct exec *t = b->right;
1290 b = cast(binode, b->left);
1300 Variables are scoped named values. We store the names in a linked list
1301 of "bindings" sorted in lexical order, and use sequential search and
1308 struct binding *next; // in lexical order
1312 This linked list is stored in the parse context so that "reduce"
1313 functions can find or add variables, and so the analysis phase can
1314 ensure that every variable gets a type.
1316 ###### parse context
1318 struct binding *varlist; // In lexical order
1320 ###### ast functions
1322 static struct binding *find_binding(struct parse_context *c, struct text s)
1324 struct binding **l = &c->varlist;
1329 (cmp = text_cmp((*l)->name, s)) < 0)
1333 n = calloc(1, sizeof(*n));
1340 Each name can be linked to multiple variables defined in different
1341 scopes. Each scope starts where the name is declared and continues
1342 until the end of the containing code block. Scopes of a given name
1343 cannot nest, so a declaration while a name is in-scope is an error.
1345 ###### binding fields
1346 struct variable *var;
1350 struct variable *previous;
1352 struct binding *name;
1353 struct exec *where_decl;// where name was declared
1354 struct exec *where_set; // where type was set
1358 When a scope closes, the values of the variables might need to be freed.
1359 This happens in the context of some `struct exec` and each `exec` will
1360 need to know which variables need to be freed when it completes.
1363 struct variable *to_free;
1365 ####### variable fields
1366 struct exec *cleanup_exec;
1367 struct variable *next_free;
1369 ####### interp exec cleanup
1372 for (v = e->to_free; v; v = v->next_free) {
1373 struct value *val = var_value(c, v);
1374 free_value(v->type, val);
1378 ###### ast functions
1379 static void variable_unlink_exec(struct variable *v)
1381 struct variable **vp;
1382 if (!v->cleanup_exec)
1384 for (vp = &v->cleanup_exec->to_free;
1385 *vp; vp = &(*vp)->next_free) {
1389 v->cleanup_exec = NULL;
1394 While the naming seems strange, we include local constants in the
1395 definition of variables. A name declared `var := value` can
1396 subsequently be changed, but a name declared `var ::= value` cannot -
1399 ###### variable fields
1402 Scopes in parallel branches can be partially merged. More
1403 specifically, if a given name is declared in both branches of an
1404 if/else then its scope is a candidate for merging. Similarly if
1405 every branch of an exhaustive switch (e.g. has an "else" clause)
1406 declares a given name, then the scopes from the branches are
1407 candidates for merging.
1409 Note that names declared inside a loop (which is only parallel to
1410 itself) are never visible after the loop. Similarly names defined in
1411 scopes which are not parallel, such as those started by `for` and
1412 `switch`, are never visible after the scope. Only variables defined in
1413 both `then` and `else` (including the implicit then after an `if`, and
1414 excluding `then` used with `for`) and in all `case`s and `else` of a
1415 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
1417 Labels, which are a bit like variables, follow different rules.
1418 Labels are not explicitly declared, but if an undeclared name appears
1419 in a context where a label is legal, that effectively declares the
1420 name as a label. The declaration remains in force (or in scope) at
1421 least to the end of the immediately containing block and conditionally
1422 in any larger containing block which does not declare the name in some
1423 other way. Importantly, the conditional scope extension happens even
1424 if the label is only used in one parallel branch of a conditional --
1425 when used in one branch it is treated as having been declared in all
1428 Merge candidates are tentatively visible beyond the end of the
1429 branching statement which creates them. If the name is used, the
1430 merge is affirmed and they become a single variable visible at the
1431 outer layer. If not - if it is redeclared first - the merge lapses.
1433 To track scopes we have an extra stack, implemented as a linked list,
1434 which roughly parallels the parse stack and which is used exclusively
1435 for scoping. When a new scope is opened, a new frame is pushed and
1436 the child-count of the parent frame is incremented. This child-count
1437 is used to distinguish between the first of a set of parallel scopes,
1438 in which declared variables must not be in scope, and subsequent
1439 branches, whether they may already be conditionally scoped.
1441 We need a total ordering of scopes so we can easily compare to variables
1442 to see if they are concurrently in scope. To achieve this we record a
1443 `scope_count` which is actually a count of both beginnings and endings
1444 of scopes. Then each variable has a record of the scope count where it
1445 enters scope, and where it leaves.
1447 To push a new frame *before* any code in the frame is parsed, we need a
1448 grammar reduction. This is most easily achieved with a grammar
1449 element which derives the empty string, and creates the new scope when
1450 it is recognised. This can be placed, for example, between a keyword
1451 like "if" and the code following it.
1455 struct scope *parent;
1459 ###### parse context
1462 struct scope *scope_stack;
1464 ###### variable fields
1465 int scope_start, scope_end;
1467 ###### ast functions
1468 static void scope_pop(struct parse_context *c)
1470 struct scope *s = c->scope_stack;
1472 c->scope_stack = s->parent;
1474 c->scope_depth -= 1;
1475 c->scope_count += 1;
1478 static void scope_push(struct parse_context *c)
1480 struct scope *s = calloc(1, sizeof(*s));
1482 c->scope_stack->child_count += 1;
1483 s->parent = c->scope_stack;
1485 c->scope_depth += 1;
1486 c->scope_count += 1;
1492 OpenScope -> ${ scope_push(c); }$
1494 Each variable records a scope depth and is in one of four states:
1496 - "in scope". This is the case between the declaration of the
1497 variable and the end of the containing block, and also between
1498 the usage with affirms a merge and the end of that block.
1500 The scope depth is not greater than the current parse context scope
1501 nest depth. When the block of that depth closes, the state will
1502 change. To achieve this, all "in scope" variables are linked
1503 together as a stack in nesting order.
1505 - "pending". The "in scope" block has closed, but other parallel
1506 scopes are still being processed. So far, every parallel block at
1507 the same level that has closed has declared the name.
1509 The scope depth is the depth of the last parallel block that
1510 enclosed the declaration, and that has closed.
1512 - "conditionally in scope". The "in scope" block and all parallel
1513 scopes have closed, and no further mention of the name has been seen.
1514 This state includes a secondary nest depth (`min_depth`) which records
1515 the outermost scope seen since the variable became conditionally in
1516 scope. If a use of the name is found, the variable becomes "in scope"
1517 and that secondary depth becomes the recorded scope depth. If the
1518 name is declared as a new variable, the old variable becomes "out of
1519 scope" and the recorded scope depth stays unchanged.
1521 - "out of scope". The variable is neither in scope nor conditionally
1522 in scope. It is permanently out of scope now and can be removed from
1523 the "in scope" stack. When a variable becomes out-of-scope it is
1524 moved to a separate list (`out_scope`) of variables which have fully
1525 known scope. This will be used at the end of each function to assign
1526 each variable a place in the stack frame.
1528 ###### variable fields
1529 int depth, min_depth;
1530 enum { OutScope, PendingScope, CondScope, InScope } scope;
1531 struct variable *in_scope;
1533 ###### parse context
1535 struct variable *in_scope;
1536 struct variable *out_scope;
1538 All variables with the same name are linked together using the
1539 'previous' link. Those variable that have been affirmatively merged all
1540 have a 'merged' pointer that points to one primary variable - the most
1541 recently declared instance. When merging variables, we need to also
1542 adjust the 'merged' pointer on any other variables that had previously
1543 been merged with the one that will no longer be primary.
1545 A variable that is no longer the most recent instance of a name may
1546 still have "pending" scope, if it might still be merged with most
1547 recent instance. These variables don't really belong in the
1548 "in_scope" list, but are not immediately removed when a new instance
1549 is found. Instead, they are detected and ignored when considering the
1550 list of in_scope names.
1552 The storage of the value of a variable will be described later. For now
1553 we just need to know that when a variable goes out of scope, it might
1554 need to be freed. For this we need to be able to find it, so assume that
1555 `var_value()` will provide that.
1557 ###### variable fields
1558 struct variable *merged;
1560 ###### ast functions
1562 static void variable_merge(struct variable *primary, struct variable *secondary)
1566 primary = primary->merged;
1568 for (v = primary->previous; v; v=v->previous)
1569 if (v == secondary || v == secondary->merged ||
1570 v->merged == secondary ||
1571 v->merged == secondary->merged) {
1572 v->scope = OutScope;
1573 v->merged = primary;
1574 if (v->scope_start < primary->scope_start)
1575 primary->scope_start = v->scope_start;
1576 if (v->scope_end > primary->scope_end)
1577 primary->scope_end = v->scope_end; // NOTEST
1578 variable_unlink_exec(v);
1582 ###### forward decls
1583 static struct value *var_value(struct parse_context *c, struct variable *v);
1585 ###### free global vars
1587 while (context.varlist) {
1588 struct binding *b = context.varlist;
1589 struct variable *v = b->var;
1590 context.varlist = b->next;
1593 struct variable *next = v->previous;
1595 if (v->global && v->frame_pos >= 0) {
1596 free_value(v->type, var_value(&context, v));
1597 if (v->depth == 0 && v->type->free == function_free)
1598 // This is a function constant
1599 free_exec(v->where_decl);
1606 #### Manipulating Bindings
1608 When a name is conditionally visible, a new declaration discards the old
1609 binding - the condition lapses. Similarly when we reach the end of a
1610 function (outermost non-global scope) any conditional scope must lapse.
1611 Conversely a usage of the name affirms the visibility and extends it to
1612 the end of the containing block - i.e. the block that contains both the
1613 original declaration and the latest usage. This is determined from
1614 `min_depth`. When a conditionally visible variable gets affirmed like
1615 this, it is also merged with other conditionally visible variables with
1618 When we parse a variable declaration we either report an error if the
1619 name is currently bound, or create a new variable at the current nest
1620 depth if the name is unbound or bound to a conditionally scoped or
1621 pending-scope variable. If the previous variable was conditionally
1622 scoped, it and its homonyms becomes out-of-scope.
1624 When we parse a variable reference (including non-declarative assignment
1625 "foo = bar") we report an error if the name is not bound or is bound to
1626 a pending-scope variable; update the scope if the name is bound to a
1627 conditionally scoped variable; or just proceed normally if the named
1628 variable is in scope.
1630 When we exit a scope, any variables bound at this level are either
1631 marked out of scope or pending-scoped, depending on whether the scope
1632 was sequential or parallel. Here a "parallel" scope means the "then"
1633 or "else" part of a conditional, or any "case" or "else" branch of a
1634 switch. Other scopes are "sequential".
1636 When exiting a parallel scope we check if there are any variables that
1637 were previously pending and are still visible. If there are, then
1638 they weren't redeclared in the most recent scope, so they cannot be
1639 merged and must become out-of-scope. If it is not the first of
1640 parallel scopes (based on `child_count`), we check that there was a
1641 previous binding that is still pending-scope. If there isn't, the new
1642 variable must now be out-of-scope.
1644 When exiting a sequential scope that immediately enclosed parallel
1645 scopes, we need to resolve any pending-scope variables. If there was
1646 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1647 we need to mark all pending-scope variable as out-of-scope. Otherwise
1648 all pending-scope variables become conditionally scoped.
1651 enum closetype { CloseSequential, CloseFunction, CloseParallel, CloseElse };
1653 ###### ast functions
1655 static struct variable *var_decl(struct parse_context *c, struct text s)
1657 struct binding *b = find_binding(c, s);
1658 struct variable *v = b->var;
1660 switch (v ? v->scope : OutScope) {
1662 /* Caller will report the error */
1666 v && v->scope == CondScope;
1668 v->scope = OutScope;
1672 v = calloc(1, sizeof(*v));
1673 v->previous = b->var;
1677 v->min_depth = v->depth = c->scope_depth;
1679 v->in_scope = c->in_scope;
1680 v->scope_start = c->scope_count;
1686 static struct variable *var_ref(struct parse_context *c, struct text s)
1688 struct binding *b = find_binding(c, s);
1689 struct variable *v = b->var;
1690 struct variable *v2;
1692 switch (v ? v->scope : OutScope) {
1695 /* Caller will report the error */
1698 /* All CondScope variables of this name need to be merged
1699 * and become InScope
1701 v->depth = v->min_depth;
1703 for (v2 = v->previous;
1704 v2 && v2->scope == CondScope;
1706 variable_merge(v, v2);
1714 static int var_refile(struct parse_context *c, struct variable *v)
1716 /* Variable just went out of scope. Add it to the out_scope
1717 * list, sorted by ->scope_start
1719 struct variable **vp = &c->out_scope;
1720 while ((*vp) && (*vp)->scope_start < v->scope_start)
1721 vp = &(*vp)->in_scope;
1727 static void var_block_close(struct parse_context *c, enum closetype ct,
1730 /* Close off all variables that are in_scope.
1731 * Some variables in c->scope may already be not-in-scope,
1732 * such as when a PendingScope variable is hidden by a new
1733 * variable with the same name.
1734 * So we check for v->name->var != v and drop them.
1735 * If we choose to make a variable OutScope, we drop it
1738 struct variable *v, **vp, *v2;
1741 for (vp = &c->in_scope;
1742 (v = *vp) && v->min_depth > c->scope_depth;
1743 (v->scope == OutScope || v->name->var != v)
1744 ? (*vp = v->in_scope, var_refile(c, v))
1745 : ( vp = &v->in_scope, 0)) {
1746 v->min_depth = c->scope_depth;
1747 if (v->name->var != v)
1748 /* This is still in scope, but we haven't just
1752 v->min_depth = c->scope_depth;
1753 if (v->scope == InScope)
1754 v->scope_end = c->scope_count;
1755 if (v->scope == InScope && e && !v->global) {
1756 /* This variable gets cleaned up when 'e' finishes */
1757 variable_unlink_exec(v);
1758 v->cleanup_exec = e;
1759 v->next_free = e->to_free;
1764 case CloseParallel: /* handle PendingScope */
1768 if (c->scope_stack->child_count == 1)
1769 /* first among parallel branches */
1770 v->scope = PendingScope;
1771 else if (v->previous &&
1772 v->previous->scope == PendingScope)
1773 /* all previous branches used name */
1774 v->scope = PendingScope;
1775 else if (v->type == Tlabel)
1776 /* Labels remain pending even when not used */
1777 v->scope = PendingScope; // UNTESTED
1779 v->scope = OutScope;
1780 if (ct == CloseElse) {
1781 /* All Pending variables with this name
1782 * are now Conditional */
1784 v2 && v2->scope == PendingScope;
1786 v2->scope = CondScope;
1790 /* Not possible as it would require
1791 * parallel scope to be nested immediately
1792 * in a parallel scope, and that never
1796 /* Not possible as we already tested for
1803 if (v->scope == CondScope)
1804 /* Condition cannot continue past end of function */
1807 case CloseSequential:
1808 if (v->type == Tlabel)
1809 v->scope = PendingScope;
1812 v->scope = OutScope;
1815 /* There was no 'else', so we can only become
1816 * conditional if we know the cases were exhaustive,
1817 * and that doesn't mean anything yet.
1818 * So only labels become conditional..
1821 v2 && v2->scope == PendingScope;
1823 if (v2->type == Tlabel)
1824 v2->scope = CondScope;
1826 v2->scope = OutScope;
1829 case OutScope: break;
1838 The value of a variable is store separately from the variable, on an
1839 analogue of a stack frame. There are (currently) two frames that can be
1840 active. A global frame which currently only stores constants, and a
1841 stacked frame which stores local variables. Each variable knows if it
1842 is global or not, and what its index into the frame is.
1844 Values in the global frame are known immediately they are relevant, so
1845 the frame needs to be reallocated as it grows so it can store those
1846 values. The local frame doesn't get values until the interpreted phase
1847 is started, so there is no need to allocate until the size is known.
1849 We initialize the `frame_pos` to an impossible value, so that we can
1850 tell if it was set or not later.
1852 ###### variable fields
1856 ###### variable init
1859 ###### parse context
1861 short global_size, global_alloc;
1863 void *global, *local;
1865 ###### forward decls
1866 static struct value *global_alloc(struct parse_context *c, struct type *t,
1867 struct variable *v, struct value *init);
1869 ###### ast functions
1871 static struct value *var_value(struct parse_context *c, struct variable *v)
1874 if (!c->local || !v->type)
1876 if (v->frame_pos + v->type->size > c->local_size) {
1877 printf("INVALID frame_pos\n"); // NOTEST
1880 return c->local + v->frame_pos;
1882 if (c->global_size > c->global_alloc) {
1883 int old = c->global_alloc;
1884 c->global_alloc = (c->global_size | 1023) + 1024;
1885 c->global = realloc(c->global, c->global_alloc);
1886 memset(c->global + old, 0, c->global_alloc - old);
1888 return c->global + v->frame_pos;
1891 static struct value *global_alloc(struct parse_context *c, struct type *t,
1892 struct variable *v, struct value *init)
1895 struct variable scratch;
1897 if (t->prepare_type)
1898 t->prepare_type(c, t, 1); // NOTEST
1900 if (c->global_size & (t->align - 1))
1901 c->global_size = (c->global_size + t->align) & ~(t->align-1); // NOTEST
1906 v->frame_pos = c->global_size;
1908 c->global_size += v->type->size;
1909 ret = var_value(c, v);
1911 memcpy(ret, init, t->size);
1917 As global values are found -- struct field initializers, labels etc --
1918 `global_alloc()` is called to record the value in the global frame.
1920 When the program is fully parsed, each function is analysed, we need to
1921 walk the list of variables local to that function and assign them an
1922 offset in the stack frame. For this we have `scope_finalize()`.
1924 We keep the stack from dense by re-using space for between variables
1925 that are not in scope at the same time. The `out_scope` list is sorted
1926 by `scope_start` and as we process a varible, we move it to an FIFO
1927 stack. For each variable we consider, we first discard any from the
1928 stack anything that went out of scope before the new variable came in.
1929 Then we place the new variable just after the one at the top of the
1932 ###### ast functions
1934 static void scope_finalize(struct parse_context *c, struct type *ft)
1936 int size = ft->function.local_size;
1937 struct variable *next = ft->function.scope;
1938 struct variable *done = NULL;
1941 struct variable *v = next;
1942 struct type *t = v->type;
1949 if (v->frame_pos >= 0)
1951 while (done && done->scope_end < v->scope_start)
1952 done = done->in_scope;
1954 pos = done->frame_pos + done->type->size;
1956 pos = ft->function.local_size;
1957 if (pos & (t->align - 1))
1958 pos = (pos + t->align) & ~(t->align-1);
1960 if (size < pos + v->type->size)
1961 size = pos + v->type->size;
1965 c->out_scope = NULL;
1966 ft->function.local_size = size;
1969 ###### free context storage
1970 free(context.global);
1972 #### Variables as executables
1974 Just as we used a `val` to wrap a value into an `exec`, we similarly
1975 need a `var` to wrap a `variable` into an exec. While each `val`
1976 contained a copy of the value, each `var` holds a link to the variable
1977 because it really is the same variable no matter where it appears.
1978 When a variable is used, we need to remember to follow the `->merged`
1979 link to find the primary instance.
1981 When a variable is declared, it may or may not be given an explicit
1982 type. We need to record which so that we can report the parsed code
1991 struct variable *var;
1994 ###### variable fields
2002 VariableDecl -> IDENTIFIER : ${ {
2003 struct variable *v = var_decl(c, $1.txt);
2004 $0 = new_pos(var, $1);
2009 v = var_ref(c, $1.txt);
2011 type_err(c, "error: variable '%v' redeclared",
2013 type_err(c, "info: this is where '%v' was first declared",
2014 v->where_decl, NULL, 0, NULL);
2017 | IDENTIFIER :: ${ {
2018 struct variable *v = var_decl(c, $1.txt);
2019 $0 = new_pos(var, $1);
2025 v = var_ref(c, $1.txt);
2027 type_err(c, "error: variable '%v' redeclared",
2029 type_err(c, "info: this is where '%v' was first declared",
2030 v->where_decl, NULL, 0, NULL);
2033 | IDENTIFIER : Type ${ {
2034 struct variable *v = var_decl(c, $1.txt);
2035 $0 = new_pos(var, $1);
2041 v->explicit_type = 1;
2043 v = var_ref(c, $1.txt);
2045 type_err(c, "error: variable '%v' redeclared",
2047 type_err(c, "info: this is where '%v' was first declared",
2048 v->where_decl, NULL, 0, NULL);
2051 | IDENTIFIER :: Type ${ {
2052 struct variable *v = var_decl(c, $1.txt);
2053 $0 = new_pos(var, $1);
2060 v->explicit_type = 1;
2062 v = var_ref(c, $1.txt);
2064 type_err(c, "error: variable '%v' redeclared",
2066 type_err(c, "info: this is where '%v' was first declared",
2067 v->where_decl, NULL, 0, NULL);
2072 Variable -> IDENTIFIER ${ {
2073 struct variable *v = var_ref(c, $1.txt);
2074 $0 = new_pos(var, $1);
2076 /* This might be a label - allocate a var just in case */
2077 v = var_decl(c, $1.txt);
2084 cast(var, $0)->var = v;
2087 ###### print exec cases
2090 struct var *v = cast(var, e);
2092 struct binding *b = v->var->name;
2093 printf("%.*s", b->name.len, b->name.txt);
2100 if (loc && loc->type == Xvar) {
2101 struct var *v = cast(var, loc);
2103 struct binding *b = v->var->name;
2104 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2106 fputs("???", stderr); // NOTEST
2108 fputs("NOTVAR", stderr);
2111 ###### propagate exec cases
2115 struct var *var = cast(var, prog);
2116 struct variable *v = var->var;
2118 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2119 return Tnone; // NOTEST
2122 if (v->constant && (rules & Rnoconstant)) {
2123 type_err(c, "error: Cannot assign to a constant: %v",
2124 prog, NULL, 0, NULL);
2125 type_err(c, "info: name was defined as a constant here",
2126 v->where_decl, NULL, 0, NULL);
2129 if (v->type == Tnone && v->where_decl == prog)
2130 type_err(c, "error: variable used but not declared: %v",
2131 prog, NULL, 0, NULL);
2132 if (v->type == NULL) {
2133 if (type && !(*perr & Efail)) {
2135 v->where_set = prog;
2140 if (!type_compat(type, v->type, rules)) {
2141 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
2142 type, rules, v->type);
2143 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2144 v->type, rules, NULL);
2151 ###### interp exec cases
2154 struct var *var = cast(var, e);
2155 struct variable *v = var->var;
2158 lrv = var_value(c, v);
2163 ###### ast functions
2165 static void free_var(struct var *v)
2170 ###### free exec cases
2171 case Xvar: free_var(cast(var, e)); break;
2176 Now that we have the shape of the interpreter in place we can add some
2177 complex types and connected them in to the data structures and the
2178 different phases of parse, analyse, print, interpret.
2180 Being "complex" the language will naturally have syntax to access
2181 specifics of objects of these types. These will fit into the grammar as
2182 "Terms" which are the things that are combined with various operators to
2183 form "Expression". Where a Term is formed by some operation on another
2184 Term, the subordinate Term will always come first, so for example a
2185 member of an array will be expressed as the Term for the array followed
2186 by an index in square brackets. The strict rule of using postfix
2187 operations makes precedence irrelevant within terms. To provide a place
2188 to put the grammar for each terms of each type, we will start out by
2189 introducing the "Term" grammar production, with contains at least a
2190 simple "Value" (to be explained later).
2194 Term -> Value ${ $0 = $<1; }$
2195 | Variable ${ $0 = $<1; }$
2198 Thus far the complex types we have are arrays and structs.
2202 Arrays can be declared by giving a size and a type, as `[size]type' so
2203 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
2204 size can be either a literal number, or a named constant. Some day an
2205 arbitrary expression will be supported.
2207 As a formal parameter to a function, the array can be declared with a
2208 new variable as the size: `name:[size::number]string`. The `size`
2209 variable is set to the size of the array and must be a constant. As
2210 `number` is the only supported type, it can be left out:
2211 `name:[size::]string`.
2213 Arrays cannot be assigned. When pointers are introduced we will also
2214 introduce array slices which can refer to part or all of an array -
2215 the assignment syntax will create a slice. For now, an array can only
2216 ever be referenced by the name it is declared with. It is likely that
2217 a "`copy`" primitive will eventually be define which can be used to
2218 make a copy of an array with controllable recursive depth.
2220 For now we have two sorts of array, those with fixed size either because
2221 it is given as a literal number or because it is a struct member (which
2222 cannot have a runtime-changing size), and those with a size that is
2223 determined at runtime - local variables with a const size. The former
2224 have their size calculated at parse time, the latter at run time.
2226 For the latter type, the `size` field of the type is the size of a
2227 pointer, and the array is reallocated every time it comes into scope.
2229 We differentiate struct fields with a const size from local variables
2230 with a const size by whether they are prepared at parse time or not.
2232 ###### type union fields
2235 int unspec; // size is unspecified - vsize must be set.
2238 struct variable *vsize;
2239 struct type *member;
2242 ###### value union fields
2243 void *array; // used if not static_size
2245 ###### value functions
2247 static void array_prepare_type(struct parse_context *c, struct type *type,
2250 struct value *vsize;
2252 if (type->array.static_size)
2254 if (type->array.unspec && parse_time)
2257 if (type->array.vsize) {
2258 vsize = var_value(c, type->array.vsize);
2262 mpz_tdiv_q(q, mpq_numref(vsize->num), mpq_denref(vsize->num));
2263 type->array.size = mpz_get_si(q);
2267 if (parse_time && type->array.member->size) {
2268 type->array.static_size = 1;
2269 type->size = type->array.size * type->array.member->size;
2270 type->align = type->array.member->align;
2274 static void array_init(struct type *type, struct value *val)
2277 void *ptr = val->ptr;
2281 if (!type->array.static_size) {
2282 val->array = calloc(type->array.size,
2283 type->array.member->size);
2286 for (i = 0; i < type->array.size; i++) {
2288 v = (void*)ptr + i * type->array.member->size;
2289 val_init(type->array.member, v);
2293 static void array_free(struct type *type, struct value *val)
2296 void *ptr = val->ptr;
2298 if (!type->array.static_size)
2300 for (i = 0; i < type->array.size; i++) {
2302 v = (void*)ptr + i * type->array.member->size;
2303 free_value(type->array.member, v);
2305 if (!type->array.static_size)
2309 static int array_compat(struct type *require, struct type *have)
2311 if (have->compat != require->compat)
2313 /* Both are arrays, so we can look at details */
2314 if (!type_compat(require->array.member, have->array.member, 0))
2316 if (have->array.unspec && require->array.unspec) {
2317 if (have->array.vsize && require->array.vsize &&
2318 have->array.vsize != require->array.vsize) // UNTESTED
2319 /* sizes might not be the same */
2320 return 0; // UNTESTED
2323 if (have->array.unspec || require->array.unspec)
2324 return 1; // UNTESTED
2325 if (require->array.vsize == NULL && have->array.vsize == NULL)
2326 return require->array.size == have->array.size;
2328 return require->array.vsize == have->array.vsize; // UNTESTED
2331 static void array_print_type(struct type *type, FILE *f)
2334 if (type->array.vsize) {
2335 struct binding *b = type->array.vsize->name;
2336 fprintf(f, "%.*s%s]", b->name.len, b->name.txt,
2337 type->array.unspec ? "::" : "");
2338 } else if (type->array.size)
2339 fprintf(f, "%d]", type->array.size);
2342 type_print(type->array.member, f);
2345 static struct type array_prototype = {
2347 .prepare_type = array_prepare_type,
2348 .print_type = array_print_type,
2349 .compat = array_compat,
2351 .size = sizeof(void*),
2352 .align = sizeof(void*),
2355 ###### declare terminals
2360 | [ NUMBER ] Type ${ {
2366 if (number_parse(num, tail, $2.txt) == 0)
2367 tok_err(c, "error: unrecognised number", &$2);
2369 tok_err(c, "error: unsupported number suffix", &$2);
2372 elements = mpz_get_ui(mpq_numref(num));
2373 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
2374 tok_err(c, "error: array size must be an integer",
2376 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
2377 tok_err(c, "error: array size is too large",
2382 $0 = t = add_anon_type(c, &array_prototype, "array[%d]", elements );
2383 t->array.size = elements;
2384 t->array.member = $<4;
2385 t->array.vsize = NULL;
2388 | [ IDENTIFIER ] Type ${ {
2389 struct variable *v = var_ref(c, $2.txt);
2392 tok_err(c, "error: name undeclared", &$2);
2393 else if (!v->constant)
2394 tok_err(c, "error: array size must be a constant", &$2);
2396 $0 = add_anon_type(c, &array_prototype, "array[%.*s]", $2.txt.len, $2.txt.txt);
2397 $0->array.member = $<4;
2399 $0->array.vsize = v;
2404 OptType -> Type ${ $0 = $<1; }$
2407 ###### formal type grammar
2409 | [ IDENTIFIER :: OptType ] Type ${ {
2410 struct variable *v = var_decl(c, $ID.txt);
2416 $0 = add_anon_type(c, &array_prototype, "array[var]");
2417 $0->array.member = $<6;
2419 $0->array.unspec = 1;
2420 $0->array.vsize = v;
2428 | Term [ Expression ] ${ {
2429 struct binode *b = new(binode);
2436 ###### print binode cases
2438 print_exec(b->left, -1, bracket);
2440 print_exec(b->right, -1, bracket);
2444 ###### propagate binode cases
2446 /* left must be an array, right must be a number,
2447 * result is the member type of the array
2449 propagate_types(b->right, c, perr, Tnum, 0);
2450 t = propagate_types(b->left, c, perr, NULL, rules & Rnoconstant);
2451 if (!t || t->compat != array_compat) {
2452 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
2455 if (!type_compat(type, t->array.member, rules)) {
2456 type_err(c, "error: have %1 but need %2", prog,
2457 t->array.member, rules, type);
2459 return t->array.member;
2463 ###### interp binode cases
2469 lleft = linterp_exec(c, b->left, <ype);
2470 right = interp_exec(c, b->right, &rtype);
2472 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
2476 if (ltype->array.static_size)
2479 ptr = *(void**)lleft;
2480 rvtype = ltype->array.member;
2481 if (i >= 0 && i < ltype->array.size)
2482 lrv = ptr + i * rvtype->size;
2484 val_init(ltype->array.member, &rv); // UNSAFE
2491 A `struct` is a data-type that contains one or more other data-types.
2492 It differs from an array in that each member can be of a different
2493 type, and they are accessed by name rather than by number. Thus you
2494 cannot choose an element by calculation, you need to know what you
2497 The language makes no promises about how a given structure will be
2498 stored in memory - it is free to rearrange fields to suit whatever
2499 criteria seems important.
2501 Structs are declared separately from program code - they cannot be
2502 declared in-line in a variable declaration like arrays can. A struct
2503 is given a name and this name is used to identify the type - the name
2504 is not prefixed by the word `struct` as it would be in C.
2506 Structs are only treated as the same if they have the same name.
2507 Simply having the same fields in the same order is not enough. This
2508 might change once we can create structure initializers from a list of
2511 Each component datum is identified much like a variable is declared,
2512 with a name, one or two colons, and a type. The type cannot be omitted
2513 as there is no opportunity to deduce the type from usage. An initial
2514 value can be given following an equals sign, so
2516 ##### Example: a struct type
2522 would declare a type called "complex" which has two number fields,
2523 each initialised to zero.
2525 Struct will need to be declared separately from the code that uses
2526 them, so we will need to be able to print out the declaration of a
2527 struct when reprinting the whole program. So a `print_type_decl` type
2528 function will be needed.
2530 ###### type union fields
2539 } *fields; // This is created when field_list is analysed.
2541 struct fieldlist *prev;
2544 } *field_list; // This is created during parsing
2547 ###### type functions
2548 void (*print_type_decl)(struct type *type, FILE *f);
2550 ###### value functions
2552 static void structure_init(struct type *type, struct value *val)
2556 for (i = 0; i < type->structure.nfields; i++) {
2558 v = (void*) val->ptr + type->structure.fields[i].offset;
2559 if (type->structure.fields[i].init)
2560 dup_value(type->structure.fields[i].type,
2561 type->structure.fields[i].init,
2564 val_init(type->structure.fields[i].type, v);
2568 static void structure_free(struct type *type, struct value *val)
2572 for (i = 0; i < type->structure.nfields; i++) {
2574 v = (void*)val->ptr + type->structure.fields[i].offset;
2575 free_value(type->structure.fields[i].type, v);
2579 static void free_fieldlist(struct fieldlist *f)
2583 free_fieldlist(f->prev);
2588 static void structure_free_type(struct type *t)
2591 for (i = 0; i < t->structure.nfields; i++)
2592 if (t->structure.fields[i].init) {
2593 free_value(t->structure.fields[i].type,
2594 t->structure.fields[i].init);
2596 free(t->structure.fields);
2597 free_fieldlist(t->structure.field_list);
2600 static void structure_prepare_type(struct parse_context *c,
2601 struct type *t, int parse_time)
2604 struct fieldlist *f;
2606 if (!parse_time || t->structure.fields)
2609 for (f = t->structure.field_list; f; f=f->prev) {
2613 if (f->f.type->prepare_type)
2614 f->f.type->prepare_type(c, f->f.type, 1);
2615 if (f->init == NULL)
2619 propagate_types(f->init, c, &perr, f->f.type, 0);
2620 } while (perr & Eretry);
2622 c->parse_error = 1; // NOTEST
2625 t->structure.nfields = cnt;
2626 t->structure.fields = calloc(cnt, sizeof(struct field));
2627 f = t->structure.field_list;
2629 int a = f->f.type->align;
2631 t->structure.fields[cnt] = f->f;
2632 if (t->size & (a-1))
2633 t->size = (t->size | (a-1)) + 1;
2634 t->structure.fields[cnt].offset = t->size;
2635 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
2639 if (f->init && !c->parse_error) {
2640 struct value vl = interp_exec(c, f->init, NULL);
2641 t->structure.fields[cnt].init =
2642 global_alloc(c, f->f.type, NULL, &vl);
2649 static struct type structure_prototype = {
2650 .init = structure_init,
2651 .free = structure_free,
2652 .free_type = structure_free_type,
2653 .print_type_decl = structure_print_type,
2654 .prepare_type = structure_prepare_type,
2668 ###### free exec cases
2670 free_exec(cast(fieldref, e)->left);
2674 ###### declare terminals
2679 | Term . IDENTIFIER ${ {
2680 struct fieldref *fr = new_pos(fieldref, $2);
2687 ###### print exec cases
2691 struct fieldref *f = cast(fieldref, e);
2692 print_exec(f->left, -1, bracket);
2693 printf(".%.*s", f->name.len, f->name.txt);
2697 ###### ast functions
2698 static int find_struct_index(struct type *type, struct text field)
2701 for (i = 0; i < type->structure.nfields; i++)
2702 if (text_cmp(type->structure.fields[i].name, field) == 0)
2707 ###### propagate exec cases
2711 struct fieldref *f = cast(fieldref, prog);
2712 struct type *st = propagate_types(f->left, c, perr, NULL, 0);
2715 type_err(c, "error: unknown type for field access", f->left, // UNTESTED
2717 else if (st->init != structure_init)
2718 type_err(c, "error: field reference attempted on %1, not a struct",
2719 f->left, st, 0, NULL);
2720 else if (f->index == -2) {
2721 f->index = find_struct_index(st, f->name);
2723 type_err(c, "error: cannot find requested field in %1",
2724 f->left, st, 0, NULL);
2726 if (f->index >= 0) {
2727 struct type *ft = st->structure.fields[f->index].type;
2728 if (!type_compat(type, ft, rules))
2729 type_err(c, "error: have %1 but need %2", prog,
2736 ###### interp exec cases
2739 struct fieldref *f = cast(fieldref, e);
2741 struct value *lleft = linterp_exec(c, f->left, <ype);
2742 lrv = (void*)lleft->ptr + ltype->structure.fields[f->index].offset;
2743 rvtype = ltype->structure.fields[f->index].type;
2747 ###### top level grammar
2748 DeclareStruct -> struct IDENTIFIER FieldBlock Newlines ${ {
2750 add_type(c, $2.txt, &structure_prototype);
2751 t->structure.field_list = $<FB;
2755 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
2756 | { SimpleFieldList } ${ $0 = $<SFL; }$
2757 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
2758 | SimpleFieldList EOL ${ $0 = $<SFL; }$
2760 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
2761 | FieldLines SimpleFieldList Newlines ${
2766 SimpleFieldList -> Field ${ $0 = $<F; }$
2767 | SimpleFieldList ; Field ${
2771 | SimpleFieldList ; ${
2774 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
2776 Field -> IDENTIFIER : Type = Expression ${ {
2777 $0 = calloc(1, sizeof(struct fieldlist));
2778 $0->f.name = $ID.txt;
2779 $0->f.type = $<Type;
2783 | IDENTIFIER : Type ${
2784 $0 = calloc(1, sizeof(struct fieldlist));
2785 $0->f.name = $ID.txt;
2786 $0->f.type = $<Type;
2789 ###### forward decls
2790 static void structure_print_type(struct type *t, FILE *f);
2792 ###### value functions
2793 static void structure_print_type(struct type *t, FILE *f)
2797 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
2799 for (i = 0; i < t->structure.nfields; i++) {
2800 struct field *fl = t->structure.fields + i;
2801 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
2802 type_print(fl->type, f);
2803 if (fl->type->print && fl->init) {
2805 if (fl->type == Tstr)
2806 fprintf(f, "\""); // UNTESTED
2807 print_value(fl->type, fl->init, f);
2808 if (fl->type == Tstr)
2809 fprintf(f, "\""); // UNTESTED
2815 ###### print type decls
2820 while (target != 0) {
2822 for (t = context.typelist; t ; t=t->next)
2823 if (!t->anon && t->print_type_decl &&
2833 t->print_type_decl(t, stdout);
2841 A function is a chunk of code which can be passed parameters and can
2842 return results. Each function has a type which includes the set of
2843 parameters and the return value. As yet these types cannot be declared
2844 separately from the function itself.
2846 The parameters can be specified either in parentheses as a ';' separated
2849 ##### Example: function 1
2851 func main(av:[ac::number]string; env:[envc::number]string)
2854 or as an indented list of one parameter per line (though each line can
2855 be a ';' separated list)
2857 ##### Example: function 2
2860 argv:[argc::number]string
2861 env:[envc::number]string
2865 In the first case a return type can follow the parentheses after a colon,
2866 in the second it is given on a line starting with the word `return`.
2868 ##### Example: functions that return
2870 func add(a:number; b:number): number
2880 Rather than returning a type, the function can specify a set of local
2881 variables to return as a struct. The values of these variables when the
2882 function exits will be provided to the caller. For this the return type
2883 is replaced with a block of result declarations, either in parentheses
2884 or bracketed by `return` and `do`.
2886 ##### Example: functions returning multiple variables
2888 func to_cartesian(rho:number; theta:number):(x:number; y:number)
2901 For constructing the lists we use a `List` binode, which will be
2902 further detailed when Expression Lists are introduced.
2904 ###### type union fields
2907 struct binode *params;
2908 struct type *return_type;
2909 struct variable *scope;
2910 int inline_result; // return value is at start of 'local'
2914 ###### value union fields
2915 struct exec *function;
2917 ###### type functions
2918 void (*check_args)(struct parse_context *c, enum prop_err *perr,
2919 struct type *require, struct exec *args);
2921 ###### value functions
2923 static void function_free(struct type *type, struct value *val)
2925 free_exec(val->function);
2926 val->function = NULL;
2929 static int function_compat(struct type *require, struct type *have)
2931 // FIXME can I do anything here yet?
2935 static void function_check_args(struct parse_context *c, enum prop_err *perr,
2936 struct type *require, struct exec *args)
2938 /* This should be 'compat', but we don't have a 'tuple' type to
2939 * hold the type of 'args'
2941 struct binode *arg = cast(binode, args);
2942 struct binode *param = require->function.params;
2945 struct var *pv = cast(var, param->left);
2947 type_err(c, "error: insufficient arguments to function.",
2948 args, NULL, 0, NULL);
2952 propagate_types(arg->left, c, perr, pv->var->type, 0);
2953 param = cast(binode, param->right);
2954 arg = cast(binode, arg->right);
2957 type_err(c, "error: too many arguments to function.",
2958 args, NULL, 0, NULL);
2961 static void function_print(struct type *type, struct value *val, FILE *f)
2963 print_exec(val->function, 1, 0);
2966 static void function_print_type_decl(struct type *type, FILE *f)
2970 for (b = type->function.params; b; b = cast(binode, b->right)) {
2971 struct variable *v = cast(var, b->left)->var;
2972 fprintf(f, "%.*s%s", v->name->name.len, v->name->name.txt,
2973 v->constant ? "::" : ":");
2974 type_print(v->type, f);
2979 if (type->function.return_type != Tnone) {
2981 if (type->function.inline_result) {
2983 struct type *t = type->function.return_type;
2985 for (i = 0; i < t->structure.nfields; i++) {
2986 struct field *fl = t->structure.fields + i;
2989 fprintf(f, "%.*s:", fl->name.len, fl->name.txt);
2990 type_print(fl->type, f);
2994 type_print(type->function.return_type, f);
2999 static void function_free_type(struct type *t)
3001 free_exec(t->function.params);
3004 static struct type function_prototype = {
3005 .size = sizeof(void*),
3006 .align = sizeof(void*),
3007 .free = function_free,
3008 .compat = function_compat,
3009 .check_args = function_check_args,
3010 .print = function_print,
3011 .print_type_decl = function_print_type_decl,
3012 .free_type = function_free_type,
3015 ###### declare terminals
3025 FuncName -> IDENTIFIER ${ {
3026 struct variable *v = var_decl(c, $1.txt);
3027 struct var *e = new_pos(var, $1);
3033 v = var_ref(c, $1.txt);
3035 type_err(c, "error: function '%v' redeclared",
3037 type_err(c, "info: this is where '%v' was first declared",
3038 v->where_decl, NULL, 0, NULL);
3044 Args -> ArgsLine NEWLINE ${ $0 = $<AL; }$
3045 | Args ArgsLine NEWLINE ${ {
3046 struct binode *b = $<AL;
3047 struct binode **bp = &b;
3049 bp = (struct binode **)&(*bp)->left;
3054 ArgsLine -> ${ $0 = NULL; }$
3055 | Varlist ${ $0 = $<1; }$
3056 | Varlist ; ${ $0 = $<1; }$
3058 Varlist -> Varlist ; ArgDecl ${
3072 ArgDecl -> IDENTIFIER : FormalType ${ {
3073 struct variable *v = var_decl(c, $1.txt);
3079 ##### Function calls
3081 A function call can appear either as an expression or as a statement.
3082 We use a new 'Funcall' binode type to link the function with a list of
3083 arguments, form with the 'List' nodes.
3085 We have already seen the "Term" which is how a function call can appear
3086 in an expression. To parse a function call into a statement we include
3087 it in the "SimpleStatement Grammar" which will be described later.
3093 | Term ( ExpressionList ) ${ {
3094 struct binode *b = new(binode);
3097 b->right = reorder_bilist($<EL);
3101 struct binode *b = new(binode);
3108 ###### SimpleStatement Grammar
3110 | Term ( ExpressionList ) ${ {
3111 struct binode *b = new(binode);
3114 b->right = reorder_bilist($<EL);
3118 ###### print binode cases
3121 do_indent(indent, "");
3122 print_exec(b->left, -1, bracket);
3124 for (b = cast(binode, b->right); b; b = cast(binode, b->right)) {
3127 print_exec(b->left, -1, bracket);
3137 ###### propagate binode cases
3140 /* Every arg must match formal parameter, and result
3141 * is return type of function
3143 struct binode *args = cast(binode, b->right);
3144 struct var *v = cast(var, b->left);
3146 if (!v->var->type || v->var->type->check_args == NULL) {
3147 type_err(c, "error: attempt to call a non-function.",
3148 prog, NULL, 0, NULL);
3151 v->var->type->check_args(c, perr, v->var->type, args);
3152 return v->var->type->function.return_type;
3155 ###### interp binode cases
3158 struct var *v = cast(var, b->left);
3159 struct type *t = v->var->type;
3160 void *oldlocal = c->local;
3161 int old_size = c->local_size;
3162 void *local = calloc(1, t->function.local_size);
3163 struct value *fbody = var_value(c, v->var);
3164 struct binode *arg = cast(binode, b->right);
3165 struct binode *param = t->function.params;
3168 struct var *pv = cast(var, param->left);
3169 struct type *vtype = NULL;
3170 struct value val = interp_exec(c, arg->left, &vtype);
3172 c->local = local; c->local_size = t->function.local_size;
3173 lval = var_value(c, pv->var);
3174 c->local = oldlocal; c->local_size = old_size;
3175 memcpy(lval, &val, vtype->size);
3176 param = cast(binode, param->right);
3177 arg = cast(binode, arg->right);
3179 c->local = local; c->local_size = t->function.local_size;
3180 if (t->function.inline_result && dtype) {
3181 _interp_exec(c, fbody->function, NULL, NULL);
3182 memcpy(dest, local, dtype->size);
3183 rvtype = ret.type = NULL;
3185 rv = interp_exec(c, fbody->function, &rvtype);
3186 c->local = oldlocal; c->local_size = old_size;
3191 ## Complex executables: statements and expressions
3193 Now that we have types and values and variables and most of the basic
3194 Terms which provide access to these, we can explore the more complex
3195 code that combine all of these to get useful work done. Specifically
3196 statements and expressions.
3198 Expressions are various combinations of Terms. We will use operator
3199 precedence to ensure correct parsing. The simplest Expression is just a
3200 Term - others will follow.
3205 Expression -> Term ${ $0 = $<Term; }$
3206 ## expression grammar
3208 ### Expressions: Conditional
3210 Our first user of the `binode` will be conditional expressions, which
3211 is a bit odd as they actually have three components. That will be
3212 handled by having 2 binodes for each expression. The conditional
3213 expression is the lowest precedence operator which is why we define it
3214 first - to start the precedence list.
3216 Conditional expressions are of the form "value `if` condition `else`
3217 other_value". They associate to the right, so everything to the right
3218 of `else` is part of an else value, while only a higher-precedence to
3219 the left of `if` is the if values. Between `if` and `else` there is no
3220 room for ambiguity, so a full conditional expression is allowed in
3226 ###### declare terminals
3230 ###### expression grammar
3232 | Expression if Expression else Expression $$ifelse ${ {
3233 struct binode *b1 = new(binode);
3234 struct binode *b2 = new(binode);
3244 ###### print binode cases
3247 b2 = cast(binode, b->right);
3248 if (bracket) printf("(");
3249 print_exec(b2->left, -1, bracket);
3251 print_exec(b->left, -1, bracket);
3253 print_exec(b2->right, -1, bracket);
3254 if (bracket) printf(")");
3257 ###### propagate binode cases
3260 /* cond must be Tbool, others must match */
3261 struct binode *b2 = cast(binode, b->right);
3264 propagate_types(b->left, c, perr, Tbool, 0);
3265 t = propagate_types(b2->left, c, perr, type, Rnolabel);
3266 t2 = propagate_types(b2->right, c, perr, type ?: t, Rnolabel);
3270 ###### interp binode cases
3273 struct binode *b2 = cast(binode, b->right);
3274 left = interp_exec(c, b->left, <ype);
3276 rv = interp_exec(c, b2->left, &rvtype); // UNTESTED
3278 rv = interp_exec(c, b2->right, &rvtype);
3284 We take a brief detour, now that we have expressions, to describe lists
3285 of expressions. These will be needed for function parameters and
3286 possibly other situations. They seem generic enough to introduce here
3287 to be used elsewhere.
3289 And ExpressionList will use the `List` type of `binode`, building up at
3290 the end. And place where they are used will probably call
3291 `reorder_bilist()` to get a more normal first/next arrangement.
3293 ###### declare terminals
3296 `List` execs have no implicit semantics, so they are never propagated or
3297 interpreted. The can be printed as a comma separate list, which is how
3298 they are parsed. Note they are also used for function formal parameter
3299 lists. In that case a separate function is used to print them.
3301 ###### print binode cases
3305 print_exec(b->left, -1, bracket);
3308 b = cast(binode, b->right);
3312 ###### propagate binode cases
3313 case List: abort(); // NOTEST
3314 ###### interp binode cases
3315 case List: abort(); // NOTEST
3320 ExpressionList -> ExpressionList , Expression ${
3333 ### Expressions: Boolean
3335 The next class of expressions to use the `binode` will be Boolean
3336 expressions. "`and then`" and "`or else`" are similar to `and` and `or`
3337 have same corresponding precendence. The difference is that they don't
3338 evaluate the second expression if not necessary.
3347 ###### declare terminals
3352 ###### expression grammar
3353 | Expression or Expression ${ {
3354 struct binode *b = new(binode);
3360 | Expression or else Expression ${ {
3361 struct binode *b = new(binode);
3368 | Expression and Expression ${ {
3369 struct binode *b = new(binode);
3375 | Expression and then Expression ${ {
3376 struct binode *b = new(binode);
3383 | not Expression ${ {
3384 struct binode *b = new(binode);
3390 ###### print binode cases
3392 if (bracket) printf("(");
3393 print_exec(b->left, -1, bracket);
3395 print_exec(b->right, -1, bracket);
3396 if (bracket) printf(")");
3399 if (bracket) printf("(");
3400 print_exec(b->left, -1, bracket);
3401 printf(" and then ");
3402 print_exec(b->right, -1, bracket);
3403 if (bracket) printf(")");
3406 if (bracket) printf("(");
3407 print_exec(b->left, -1, bracket);
3409 print_exec(b->right, -1, bracket);
3410 if (bracket) printf(")");
3413 if (bracket) printf("(");
3414 print_exec(b->left, -1, bracket);
3415 printf(" or else ");
3416 print_exec(b->right, -1, bracket);
3417 if (bracket) printf(")");
3420 if (bracket) printf("(");
3422 print_exec(b->right, -1, bracket);
3423 if (bracket) printf(")");
3426 ###### propagate binode cases
3432 /* both must be Tbool, result is Tbool */
3433 propagate_types(b->left, c, perr, Tbool, 0);
3434 propagate_types(b->right, c, perr, Tbool, 0);
3435 if (type && type != Tbool)
3436 type_err(c, "error: %1 operation found where %2 expected", prog,
3440 ###### interp binode cases
3442 rv = interp_exec(c, b->left, &rvtype);
3443 right = interp_exec(c, b->right, &rtype);
3444 rv.bool = rv.bool && right.bool;
3447 rv = interp_exec(c, b->left, &rvtype);
3449 rv = interp_exec(c, b->right, NULL);
3452 rv = interp_exec(c, b->left, &rvtype);
3453 right = interp_exec(c, b->right, &rtype);
3454 rv.bool = rv.bool || right.bool;
3457 rv = interp_exec(c, b->left, &rvtype);
3459 rv = interp_exec(c, b->right, NULL);
3462 rv = interp_exec(c, b->right, &rvtype);
3466 ### Expressions: Comparison
3468 Of slightly higher precedence that Boolean expressions are Comparisons.
3469 A comparison takes arguments of any comparable type, but the two types
3472 To simplify the parsing we introduce an `eop` which can record an
3473 expression operator, and the `CMPop` non-terminal will match one of them.
3480 ###### ast functions
3481 static void free_eop(struct eop *e)
3495 ###### declare terminals
3496 $LEFT < > <= >= == != CMPop
3498 ###### expression grammar
3499 | Expression CMPop Expression ${ {
3500 struct binode *b = new(binode);
3510 CMPop -> < ${ $0.op = Less; }$
3511 | > ${ $0.op = Gtr; }$
3512 | <= ${ $0.op = LessEq; }$
3513 | >= ${ $0.op = GtrEq; }$
3514 | == ${ $0.op = Eql; }$
3515 | != ${ $0.op = NEql; }$
3517 ###### print binode cases
3525 if (bracket) printf("(");
3526 print_exec(b->left, -1, bracket);
3528 case Less: printf(" < "); break;
3529 case LessEq: printf(" <= "); break;
3530 case Gtr: printf(" > "); break;
3531 case GtrEq: printf(" >= "); break;
3532 case Eql: printf(" == "); break;
3533 case NEql: printf(" != "); break;
3534 default: abort(); // NOTEST
3536 print_exec(b->right, -1, bracket);
3537 if (bracket) printf(")");
3540 ###### propagate binode cases
3547 /* Both must match but not be labels, result is Tbool */
3548 t = propagate_types(b->left, c, perr, NULL, Rnolabel);
3550 propagate_types(b->right, c, perr, t, 0);
3552 t = propagate_types(b->right, c, perr, NULL, Rnolabel); // UNTESTED
3554 t = propagate_types(b->left, c, perr, t, 0); // UNTESTED
3556 if (!type_compat(type, Tbool, 0))
3557 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
3558 Tbool, rules, type);
3561 ###### interp binode cases
3570 left = interp_exec(c, b->left, <ype);
3571 right = interp_exec(c, b->right, &rtype);
3572 cmp = value_cmp(ltype, rtype, &left, &right);
3575 case Less: rv.bool = cmp < 0; break;
3576 case LessEq: rv.bool = cmp <= 0; break;
3577 case Gtr: rv.bool = cmp > 0; break;
3578 case GtrEq: rv.bool = cmp >= 0; break;
3579 case Eql: rv.bool = cmp == 0; break;
3580 case NEql: rv.bool = cmp != 0; break;
3581 default: rv.bool = 0; break; // NOTEST
3586 ### Expressions: Arithmetic etc.
3588 The remaining expressions with the highest precedence are arithmetic,
3589 string concatenation, and string conversion. String concatenation
3590 (`++`) has the same precedence as multiplication and division, but lower
3593 String conversion is a temporary feature until I get a better type
3594 system. `$` is a prefix operator which expects a string and returns
3597 `+` and `-` are both infix and prefix operations (where they are
3598 absolute value and negation). These have different operator names.
3600 We also have a 'Bracket' operator which records where parentheses were
3601 found. This makes it easy to reproduce these when printing. Possibly I
3602 should only insert brackets were needed for precedence. Putting
3603 parentheses around an expression converts it into a Term,
3613 ###### declare terminals
3619 ###### expression grammar
3620 | Expression Eop Expression ${ {
3621 struct binode *b = new(binode);
3628 | Expression Top Expression ${ {
3629 struct binode *b = new(binode);
3636 | Uop Expression ${ {
3637 struct binode *b = new(binode);
3645 | ( Expression ) ${ {
3646 struct binode *b = new_pos(binode, $1);
3655 Eop -> + ${ $0.op = Plus; }$
3656 | - ${ $0.op = Minus; }$
3658 Uop -> + ${ $0.op = Absolute; }$
3659 | - ${ $0.op = Negate; }$
3660 | $ ${ $0.op = StringConv; }$
3662 Top -> * ${ $0.op = Times; }$
3663 | / ${ $0.op = Divide; }$
3664 | % ${ $0.op = Rem; }$
3665 | ++ ${ $0.op = Concat; }$
3667 ###### print binode cases
3674 if (bracket) printf("(");
3675 print_exec(b->left, indent, bracket);
3677 case Plus: fputs(" + ", stdout); break;
3678 case Minus: fputs(" - ", stdout); break;
3679 case Times: fputs(" * ", stdout); break;
3680 case Divide: fputs(" / ", stdout); break;
3681 case Rem: fputs(" % ", stdout); break;
3682 case Concat: fputs(" ++ ", stdout); break;
3683 default: abort(); // NOTEST
3685 print_exec(b->right, indent, bracket);
3686 if (bracket) printf(")");
3691 if (bracket) printf("(");
3693 case Absolute: fputs("+", stdout); break;
3694 case Negate: fputs("-", stdout); break;
3695 case StringConv: fputs("$", stdout); break;
3696 default: abort(); // NOTEST
3698 print_exec(b->right, indent, bracket);
3699 if (bracket) printf(")");
3703 print_exec(b->right, indent, bracket);
3707 ###### propagate binode cases
3713 /* both must be numbers, result is Tnum */
3716 /* as propagate_types ignores a NULL,
3717 * unary ops fit here too */
3718 propagate_types(b->left, c, perr, Tnum, 0);
3719 propagate_types(b->right, c, perr, Tnum, 0);
3720 if (!type_compat(type, Tnum, 0))
3721 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
3726 /* both must be Tstr, result is Tstr */
3727 propagate_types(b->left, c, perr, Tstr, 0);
3728 propagate_types(b->right, c, perr, Tstr, 0);
3729 if (!type_compat(type, Tstr, 0))
3730 type_err(c, "error: Concat returns %1 but %2 expected", prog,
3735 /* op must be string, result is number */
3736 propagate_types(b->left, c, perr, Tstr, 0);
3737 if (!type_compat(type, Tnum, 0))
3738 type_err(c, // UNTESTED
3739 "error: Can only convert string to number, not %1",
3740 prog, type, 0, NULL);
3744 return propagate_types(b->right, c, perr, type, 0);
3746 ###### interp binode cases
3749 rv = interp_exec(c, b->left, &rvtype);
3750 right = interp_exec(c, b->right, &rtype);
3751 mpq_add(rv.num, rv.num, right.num);
3754 rv = interp_exec(c, b->left, &rvtype);
3755 right = interp_exec(c, b->right, &rtype);
3756 mpq_sub(rv.num, rv.num, right.num);
3759 rv = interp_exec(c, b->left, &rvtype);
3760 right = interp_exec(c, b->right, &rtype);
3761 mpq_mul(rv.num, rv.num, right.num);
3764 rv = interp_exec(c, b->left, &rvtype);
3765 right = interp_exec(c, b->right, &rtype);
3766 mpq_div(rv.num, rv.num, right.num);
3771 left = interp_exec(c, b->left, <ype);
3772 right = interp_exec(c, b->right, &rtype);
3773 mpz_init(l); mpz_init(r); mpz_init(rem);
3774 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
3775 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
3776 mpz_tdiv_r(rem, l, r);
3777 val_init(Tnum, &rv);
3778 mpq_set_z(rv.num, rem);
3779 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
3784 rv = interp_exec(c, b->right, &rvtype);
3785 mpq_neg(rv.num, rv.num);
3788 rv = interp_exec(c, b->right, &rvtype);
3789 mpq_abs(rv.num, rv.num);
3792 rv = interp_exec(c, b->right, &rvtype);
3795 left = interp_exec(c, b->left, <ype);
3796 right = interp_exec(c, b->right, &rtype);
3798 rv.str = text_join(left.str, right.str);
3801 right = interp_exec(c, b->right, &rvtype);
3805 struct text tx = right.str;
3808 if (tx.txt[0] == '-') {
3809 neg = 1; // UNTESTED
3810 tx.txt++; // UNTESTED
3811 tx.len--; // UNTESTED
3813 if (number_parse(rv.num, tail, tx) == 0)
3814 mpq_init(rv.num); // UNTESTED
3816 mpq_neg(rv.num, rv.num); // UNTESTED
3818 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt); // UNTESTED
3822 ###### value functions
3824 static struct text text_join(struct text a, struct text b)
3827 rv.len = a.len + b.len;
3828 rv.txt = malloc(rv.len);
3829 memcpy(rv.txt, a.txt, a.len);
3830 memcpy(rv.txt+a.len, b.txt, b.len);
3834 ### Blocks, Statements, and Statement lists.
3836 Now that we have expressions out of the way we need to turn to
3837 statements. There are simple statements and more complex statements.
3838 Simple statements do not contain (syntactic) newlines, complex statements do.
3840 Statements often come in sequences and we have corresponding simple
3841 statement lists and complex statement lists.
3842 The former comprise only simple statements separated by semicolons.
3843 The later comprise complex statements and simple statement lists. They are
3844 separated by newlines. Thus the semicolon is only used to separate
3845 simple statements on the one line. This may be overly restrictive,
3846 but I'm not sure I ever want a complex statement to share a line with
3849 Note that a simple statement list can still use multiple lines if
3850 subsequent lines are indented, so
3852 ###### Example: wrapped simple statement list
3857 is a single simple statement list. This might allow room for
3858 confusion, so I'm not set on it yet.
3860 A simple statement list needs no extra syntax. A complex statement
3861 list has two syntactic forms. It can be enclosed in braces (much like
3862 C blocks), or it can be introduced by an indent and continue until an
3863 unindented newline (much like Python blocks). With this extra syntax
3864 it is referred to as a block.
3866 Note that a block does not have to include any newlines if it only
3867 contains simple statements. So both of:
3869 if condition: a=b; d=f
3871 if condition { a=b; print f }
3875 In either case the list is constructed from a `binode` list with
3876 `Block` as the operator. When parsing the list it is most convenient
3877 to append to the end, so a list is a list and a statement. When using
3878 the list it is more convenient to consider a list to be a statement
3879 and a list. So we need a function to re-order a list.
3880 `reorder_bilist` serves this purpose.
3882 The only stand-alone statement we introduce at this stage is `pass`
3883 which does nothing and is represented as a `NULL` pointer in a `Block`
3884 list. Other stand-alone statements will follow once the infrastructure
3887 As many statements will use binodes, we declare a binode pointer 'b' in
3888 the common header for all reductions to use.
3890 ###### Parser: reduce
3901 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3902 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3903 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3904 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3905 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3907 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3908 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3909 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3910 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3911 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3913 UseBlock -> { OpenScope IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3914 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3915 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3917 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3918 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3919 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3920 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3921 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3923 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
3925 ComplexStatements -> ComplexStatements ComplexStatement ${
3935 | ComplexStatement ${
3947 ComplexStatement -> SimpleStatements Newlines ${
3948 $0 = reorder_bilist($<SS);
3950 | SimpleStatements ; Newlines ${
3951 $0 = reorder_bilist($<SS);
3953 ## ComplexStatement Grammar
3956 SimpleStatements -> SimpleStatements ; SimpleStatement ${
3962 | SimpleStatement ${
3971 SimpleStatement -> pass ${ $0 = NULL; }$
3972 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
3973 ## SimpleStatement Grammar
3975 ###### print binode cases
3979 if (b->left == NULL) // UNTESTED
3980 printf("pass"); // UNTESTED
3982 print_exec(b->left, indent, bracket); // UNTESTED
3983 if (b->right) { // UNTESTED
3984 printf("; "); // UNTESTED
3985 print_exec(b->right, indent, bracket); // UNTESTED
3988 // block, one per line
3989 if (b->left == NULL)
3990 do_indent(indent, "pass\n");
3992 print_exec(b->left, indent, bracket);
3994 print_exec(b->right, indent, bracket);
3998 ###### propagate binode cases
4001 /* If any statement returns something other than Tnone
4002 * or Tbool then all such must return same type.
4003 * As each statement may be Tnone or something else,
4004 * we must always pass NULL (unknown) down, otherwise an incorrect
4005 * error might occur. We never return Tnone unless it is
4010 for (e = b; e; e = cast(binode, e->right)) {
4011 t = propagate_types(e->left, c, perr, NULL, rules);
4012 if ((rules & Rboolok) && (t == Tbool || t == Tnone))
4014 if (t == Tnone && e->right)
4015 /* Only the final statement *must* return a value
4023 type_err(c, "error: expected %1%r, found %2",
4024 e->left, type, rules, t);
4030 ###### interp binode cases
4032 while (rvtype == Tnone &&
4035 rv = interp_exec(c, b->left, &rvtype);
4036 b = cast(binode, b->right);
4040 ### The Print statement
4042 `print` is a simple statement that takes a comma-separated list of
4043 expressions and prints the values separated by spaces and terminated
4044 by a newline. No control of formatting is possible.
4046 `print` uses `ExpressionList` to collect the expressions and stores them
4047 on the left side of a `Print` binode unlessthere is a trailing comma
4048 when the list is stored on the `right` side and no trailing newline is
4054 ##### declare terminals
4057 ###### SimpleStatement Grammar
4059 | print ExpressionList ${
4060 $0 = b = new(binode);
4063 b->left = reorder_bilist($<EL);
4065 | print ExpressionList , ${ {
4066 $0 = b = new(binode);
4068 b->right = reorder_bilist($<EL);
4072 $0 = b = new(binode);
4078 ###### print binode cases
4081 do_indent(indent, "print");
4083 print_exec(b->right, -1, bracket);
4086 print_exec(b->left, -1, bracket);
4091 ###### propagate binode cases
4094 /* don't care but all must be consistent */
4096 b = cast(binode, b->left);
4098 b = cast(binode, b->right);
4100 propagate_types(b->left, c, perr, NULL, Rnolabel);
4101 b = cast(binode, b->right);
4105 ###### interp binode cases
4109 struct binode *b2 = cast(binode, b->left);
4111 b2 = cast(binode, b->right);
4112 for (; b2; b2 = cast(binode, b2->right)) {
4113 left = interp_exec(c, b2->left, <ype);
4114 print_value(ltype, &left, stdout);
4115 free_value(ltype, &left);
4119 if (b->right == NULL)
4125 ###### Assignment statement
4127 An assignment will assign a value to a variable, providing it hasn't
4128 been declared as a constant. The analysis phase ensures that the type
4129 will be correct so the interpreter just needs to perform the
4130 calculation. There is a form of assignment which declares a new
4131 variable as well as assigning a value. If a name is assigned before
4132 it is declared, and error will be raised as the name is created as
4133 `Tlabel` and it is illegal to assign to such names.
4139 ###### declare terminals
4142 ###### SimpleStatement Grammar
4143 | Term = Expression ${
4144 $0 = b= new(binode);
4149 | VariableDecl = Expression ${
4150 $0 = b= new(binode);
4157 if ($1->var->where_set == NULL) {
4159 "Variable declared with no type or value: %v",
4163 $0 = b = new(binode);
4170 ###### print binode cases
4173 do_indent(indent, "");
4174 print_exec(b->left, indent, bracket);
4176 print_exec(b->right, indent, bracket);
4183 struct variable *v = cast(var, b->left)->var;
4184 do_indent(indent, "");
4185 print_exec(b->left, indent, bracket);
4186 if (cast(var, b->left)->var->constant) {
4188 if (v->explicit_type) {
4189 type_print(v->type, stdout);
4194 if (v->explicit_type) {
4195 type_print(v->type, stdout);
4201 print_exec(b->right, indent, bracket);
4208 ###### propagate binode cases
4212 /* Both must match and not be labels,
4213 * Type must support 'dup',
4214 * For Assign, left must not be constant.
4217 t = propagate_types(b->left, c, perr, NULL,
4218 Rnolabel | (b->op == Assign ? Rnoconstant : 0));
4223 if (propagate_types(b->right, c, perr, t, 0) != t)
4224 if (b->left->type == Xvar)
4225 type_err(c, "info: variable '%v' was set as %1 here.",
4226 cast(var, b->left)->var->where_set, t, rules, NULL);
4228 t = propagate_types(b->right, c, perr, NULL, Rnolabel);
4230 propagate_types(b->left, c, perr, t,
4231 (b->op == Assign ? Rnoconstant : 0));
4233 if (t && t->dup == NULL && t->name.txt[0] != ' ') // HACK
4234 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
4239 ###### interp binode cases
4242 lleft = linterp_exec(c, b->left, <ype);
4244 dinterp_exec(c, b->right, lleft, ltype, 1);
4250 struct variable *v = cast(var, b->left)->var;
4253 val = var_value(c, v);
4254 if (v->type->prepare_type)
4255 v->type->prepare_type(c, v->type, 0);
4257 dinterp_exec(c, b->right, val, v->type, 0);
4259 val_init(v->type, val);
4263 ### The `use` statement
4265 The `use` statement is the last "simple" statement. It is needed when a
4266 statement block can return a value. This includes the body of a
4267 function which has a return type, and the "condition" code blocks in
4268 `if`, `while`, and `switch` statements.
4273 ###### declare terminals
4276 ###### SimpleStatement Grammar
4278 $0 = b = new_pos(binode, $1);
4281 if (b->right->type == Xvar) {
4282 struct var *v = cast(var, b->right);
4283 if (v->var->type == Tnone) {
4284 /* Convert this to a label */
4287 v->var->type = Tlabel;
4288 val = global_alloc(c, Tlabel, v->var, NULL);
4294 ###### print binode cases
4297 do_indent(indent, "use ");
4298 print_exec(b->right, -1, bracket);
4303 ###### propagate binode cases
4306 /* result matches value */
4307 return propagate_types(b->right, c, perr, type, 0);
4309 ###### interp binode cases
4312 rv = interp_exec(c, b->right, &rvtype);
4315 ### The Conditional Statement
4317 This is the biggy and currently the only complex statement. This
4318 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
4319 It is comprised of a number of parts, all of which are optional though
4320 set combinations apply. Each part is (usually) a key word (`then` is
4321 sometimes optional) followed by either an expression or a code block,
4322 except the `casepart` which is a "key word and an expression" followed
4323 by a code block. The code-block option is valid for all parts and,
4324 where an expression is also allowed, the code block can use the `use`
4325 statement to report a value. If the code block does not report a value
4326 the effect is similar to reporting `True`.
4328 The `else` and `case` parts, as well as `then` when combined with
4329 `if`, can contain a `use` statement which will apply to some
4330 containing conditional statement. `for` parts, `do` parts and `then`
4331 parts used with `for` can never contain a `use`, except in some
4332 subordinate conditional statement.
4334 If there is a `forpart`, it is executed first, only once.
4335 If there is a `dopart`, then it is executed repeatedly providing
4336 always that the `condpart` or `cond`, if present, does not return a non-True
4337 value. `condpart` can fail to return any value if it simply executes
4338 to completion. This is treated the same as returning `True`.
4340 If there is a `thenpart` it will be executed whenever the `condpart`
4341 or `cond` returns True (or does not return any value), but this will happen
4342 *after* `dopart` (when present).
4344 If `elsepart` is present it will be executed at most once when the
4345 condition returns `False` or some value that isn't `True` and isn't
4346 matched by any `casepart`. If there are any `casepart`s, they will be
4347 executed when the condition returns a matching value.
4349 The particular sorts of values allowed in case parts has not yet been
4350 determined in the language design, so nothing is prohibited.
4352 The various blocks in this complex statement potentially provide scope
4353 for variables as described earlier. Each such block must include the
4354 "OpenScope" nonterminal before parsing the block, and must call
4355 `var_block_close()` when closing the block.
4357 The code following "`if`", "`switch`" and "`for`" does not get its own
4358 scope, but is in a scope covering the whole statement, so names
4359 declared there cannot be redeclared elsewhere. Similarly the
4360 condition following "`while`" is in a scope the covers the body
4361 ("`do`" part) of the loop, and which does not allow conditional scope
4362 extension. Code following "`then`" (both looping and non-looping),
4363 "`else`" and "`case`" each get their own local scope.
4365 The type requirements on the code block in a `whilepart` are quite
4366 unusal. It is allowed to return a value of some identifiable type, in
4367 which case the loop aborts and an appropriate `casepart` is run, or it
4368 can return a Boolean, in which case the loop either continues to the
4369 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
4370 This is different both from the `ifpart` code block which is expected to
4371 return a Boolean, or the `switchpart` code block which is expected to
4372 return the same type as the casepart values. The correct analysis of
4373 the type of the `whilepart` code block is the reason for the
4374 `Rboolok` flag which is passed to `propagate_types()`.
4376 The `cond_statement` cannot fit into a `binode` so a new `exec` is
4377 defined. As there are two scopes which cover multiple parts - one for
4378 the whole statement and one for "while" and "do" - and as we will use
4379 the 'struct exec' to track scopes, we actually need two new types of
4380 exec. One is a `binode` for the looping part, the rest is the
4381 `cond_statement`. The `cond_statement` will use an auxilliary `struct
4382 casepart` to track a list of case parts.
4393 struct exec *action;
4394 struct casepart *next;
4396 struct cond_statement {
4398 struct exec *forpart, *condpart, *thenpart, *elsepart;
4399 struct binode *looppart;
4400 struct casepart *casepart;
4403 ###### ast functions
4405 static void free_casepart(struct casepart *cp)
4409 free_exec(cp->value);
4410 free_exec(cp->action);
4417 static void free_cond_statement(struct cond_statement *s)
4421 free_exec(s->forpart);
4422 free_exec(s->condpart);
4423 free_exec(s->looppart);
4424 free_exec(s->thenpart);
4425 free_exec(s->elsepart);
4426 free_casepart(s->casepart);
4430 ###### free exec cases
4431 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
4433 ###### ComplexStatement Grammar
4434 | CondStatement ${ $0 = $<1; }$
4436 ###### declare terminals
4437 $TERM for then while do
4444 // A CondStatement must end with EOL, as does CondSuffix and
4446 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
4447 // may or may not end with EOL
4448 // WhilePart and IfPart include an appropriate Suffix
4450 // ForPart, SwitchPart, and IfPart open scopes, o we have to close
4451 // them. WhilePart opens and closes its own scope.
4452 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
4455 $0->thenpart = $<TP;
4456 $0->looppart = $<WP;
4457 var_block_close(c, CloseSequential, $0);
4459 | ForPart OptNL WhilePart CondSuffix ${
4462 $0->looppart = $<WP;
4463 var_block_close(c, CloseSequential, $0);
4465 | WhilePart CondSuffix ${
4467 $0->looppart = $<WP;
4469 | SwitchPart OptNL CasePart CondSuffix ${
4471 $0->condpart = $<SP;
4472 $CP->next = $0->casepart;
4473 $0->casepart = $<CP;
4474 var_block_close(c, CloseSequential, $0);
4476 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
4478 $0->condpart = $<SP;
4479 $CP->next = $0->casepart;
4480 $0->casepart = $<CP;
4481 var_block_close(c, CloseSequential, $0);
4483 | IfPart IfSuffix ${
4485 $0->condpart = $IP.condpart; $IP.condpart = NULL;
4486 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
4487 // This is where we close an "if" statement
4488 var_block_close(c, CloseSequential, $0);
4491 CondSuffix -> IfSuffix ${
4494 | Newlines CasePart CondSuffix ${
4496 $CP->next = $0->casepart;
4497 $0->casepart = $<CP;
4499 | CasePart CondSuffix ${
4501 $CP->next = $0->casepart;
4502 $0->casepart = $<CP;
4505 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
4506 | Newlines ElsePart ${ $0 = $<EP; }$
4507 | ElsePart ${$0 = $<EP; }$
4509 ElsePart -> else OpenBlock Newlines ${
4510 $0 = new(cond_statement);
4511 $0->elsepart = $<OB;
4512 var_block_close(c, CloseElse, $0->elsepart);
4514 | else OpenScope CondStatement ${
4515 $0 = new(cond_statement);
4516 $0->elsepart = $<CS;
4517 var_block_close(c, CloseElse, $0->elsepart);
4521 CasePart -> case Expression OpenScope ColonBlock ${
4522 $0 = calloc(1,sizeof(struct casepart));
4525 var_block_close(c, CloseParallel, $0->action);
4529 // These scopes are closed in CondStatement
4530 ForPart -> for OpenBlock ${
4534 ThenPart -> then OpenBlock ${
4536 var_block_close(c, CloseSequential, $0);
4540 // This scope is closed in CondStatement
4541 WhilePart -> while UseBlock OptNL do OpenBlock ${
4546 var_block_close(c, CloseSequential, $0->right);
4547 var_block_close(c, CloseSequential, $0);
4549 | while OpenScope Expression OpenScope ColonBlock ${
4554 var_block_close(c, CloseSequential, $0->right);
4555 var_block_close(c, CloseSequential, $0);
4559 IfPart -> if UseBlock OptNL then OpenBlock ${
4562 var_block_close(c, CloseParallel, $0.thenpart);
4564 | if OpenScope Expression OpenScope ColonBlock ${
4567 var_block_close(c, CloseParallel, $0.thenpart);
4569 | if OpenScope Expression OpenScope OptNL then Block ${
4572 var_block_close(c, CloseParallel, $0.thenpart);
4576 // This scope is closed in CondStatement
4577 SwitchPart -> switch OpenScope Expression ${
4580 | switch UseBlock ${
4584 ###### print binode cases
4586 if (b->left && b->left->type == Xbinode &&
4587 cast(binode, b->left)->op == Block) {
4589 do_indent(indent, "while {\n");
4591 do_indent(indent, "while\n");
4592 print_exec(b->left, indent+1, bracket);
4594 do_indent(indent, "} do {\n");
4596 do_indent(indent, "do\n");
4597 print_exec(b->right, indent+1, bracket);
4599 do_indent(indent, "}\n");
4601 do_indent(indent, "while ");
4602 print_exec(b->left, 0, bracket);
4607 print_exec(b->right, indent+1, bracket);
4609 do_indent(indent, "}\n");
4613 ###### print exec cases
4615 case Xcond_statement:
4617 struct cond_statement *cs = cast(cond_statement, e);
4618 struct casepart *cp;
4620 do_indent(indent, "for");
4621 if (bracket) printf(" {\n"); else printf("\n");
4622 print_exec(cs->forpart, indent+1, bracket);
4625 do_indent(indent, "} then {\n");
4627 do_indent(indent, "then\n");
4628 print_exec(cs->thenpart, indent+1, bracket);
4630 if (bracket) do_indent(indent, "}\n");
4633 print_exec(cs->looppart, indent, bracket);
4637 do_indent(indent, "switch");
4639 do_indent(indent, "if");
4640 if (cs->condpart && cs->condpart->type == Xbinode &&
4641 cast(binode, cs->condpart)->op == Block) {
4646 print_exec(cs->condpart, indent+1, bracket);
4648 do_indent(indent, "}\n");
4650 do_indent(indent, "then\n");
4651 print_exec(cs->thenpart, indent+1, bracket);
4655 print_exec(cs->condpart, 0, bracket);
4661 print_exec(cs->thenpart, indent+1, bracket);
4663 do_indent(indent, "}\n");
4668 for (cp = cs->casepart; cp; cp = cp->next) {
4669 do_indent(indent, "case ");
4670 print_exec(cp->value, -1, 0);
4675 print_exec(cp->action, indent+1, bracket);
4677 do_indent(indent, "}\n");
4680 do_indent(indent, "else");
4685 print_exec(cs->elsepart, indent+1, bracket);
4687 do_indent(indent, "}\n");
4692 ###### propagate binode cases
4694 t = propagate_types(b->right, c, perr, Tnone, 0);
4695 if (!type_compat(Tnone, t, 0))
4696 *perr |= Efail; // UNTESTED
4697 return propagate_types(b->left, c, perr, type, rules);
4699 ###### propagate exec cases
4700 case Xcond_statement:
4702 // forpart and looppart->right must return Tnone
4703 // thenpart must return Tnone if there is a loopart,
4704 // otherwise it is like elsepart.
4706 // be bool if there is no casepart
4707 // match casepart->values if there is a switchpart
4708 // either be bool or match casepart->value if there
4710 // elsepart and casepart->action must match the return type
4711 // expected of this statement.
4712 struct cond_statement *cs = cast(cond_statement, prog);
4713 struct casepart *cp;
4715 t = propagate_types(cs->forpart, c, perr, Tnone, 0);
4716 if (!type_compat(Tnone, t, 0))
4717 *perr |= Efail; // UNTESTED
4720 t = propagate_types(cs->thenpart, c, perr, Tnone, 0);
4721 if (!type_compat(Tnone, t, 0))
4722 *perr |= Efail; // UNTESTED
4724 if (cs->casepart == NULL) {
4725 propagate_types(cs->condpart, c, perr, Tbool, 0);
4726 propagate_types(cs->looppart, c, perr, Tbool, 0);
4728 /* Condpart must match case values, with bool permitted */
4730 for (cp = cs->casepart;
4731 cp && !t; cp = cp->next)
4732 t = propagate_types(cp->value, c, perr, NULL, 0);
4733 if (!t && cs->condpart)
4734 t = propagate_types(cs->condpart, c, perr, NULL, Rboolok); // UNTESTED
4735 if (!t && cs->looppart)
4736 t = propagate_types(cs->looppart, c, perr, NULL, Rboolok); // UNTESTED
4737 // Now we have a type (I hope) push it down
4739 for (cp = cs->casepart; cp; cp = cp->next)
4740 propagate_types(cp->value, c, perr, t, 0);
4741 propagate_types(cs->condpart, c, perr, t, Rboolok);
4742 propagate_types(cs->looppart, c, perr, t, Rboolok);
4745 // (if)then, else, and case parts must return expected type.
4746 if (!cs->looppart && !type)
4747 type = propagate_types(cs->thenpart, c, perr, NULL, rules);
4749 type = propagate_types(cs->elsepart, c, perr, NULL, rules);
4750 for (cp = cs->casepart;
4752 cp = cp->next) // UNTESTED
4753 type = propagate_types(cp->action, c, perr, NULL, rules); // UNTESTED
4756 propagate_types(cs->thenpart, c, perr, type, rules);
4757 propagate_types(cs->elsepart, c, perr, type, rules);
4758 for (cp = cs->casepart; cp ; cp = cp->next)
4759 propagate_types(cp->action, c, perr, type, rules);
4765 ###### interp binode cases
4767 // This just performs one iterration of the loop
4768 rv = interp_exec(c, b->left, &rvtype);
4769 if (rvtype == Tnone ||
4770 (rvtype == Tbool && rv.bool != 0))
4771 // rvtype is Tnone or Tbool, doesn't need to be freed
4772 interp_exec(c, b->right, NULL);
4775 ###### interp exec cases
4776 case Xcond_statement:
4778 struct value v, cnd;
4779 struct type *vtype, *cndtype;
4780 struct casepart *cp;
4781 struct cond_statement *cs = cast(cond_statement, e);
4784 interp_exec(c, cs->forpart, NULL);
4786 while ((cnd = interp_exec(c, cs->looppart, &cndtype)),
4787 cndtype == Tnone || (cndtype == Tbool && cnd.bool != 0))
4788 interp_exec(c, cs->thenpart, NULL);
4790 cnd = interp_exec(c, cs->condpart, &cndtype);
4791 if ((cndtype == Tnone ||
4792 (cndtype == Tbool && cnd.bool != 0))) {
4793 // cnd is Tnone or Tbool, doesn't need to be freed
4794 rv = interp_exec(c, cs->thenpart, &rvtype);
4795 // skip else (and cases)
4799 for (cp = cs->casepart; cp; cp = cp->next) {
4800 v = interp_exec(c, cp->value, &vtype);
4801 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
4802 free_value(vtype, &v);
4803 free_value(cndtype, &cnd);
4804 rv = interp_exec(c, cp->action, &rvtype);
4807 free_value(vtype, &v);
4809 free_value(cndtype, &cnd);
4811 rv = interp_exec(c, cs->elsepart, &rvtype);
4818 ### Top level structure
4820 All the language elements so far can be used in various places. Now
4821 it is time to clarify what those places are.
4823 At the top level of a file there will be a number of declarations.
4824 Many of the things that can be declared haven't been described yet,
4825 such as functions, procedures, imports, and probably more.
4826 For now there are two sorts of things that can appear at the top
4827 level. They are predefined constants, `struct` types, and the `main`
4828 function. While the syntax will allow the `main` function to appear
4829 multiple times, that will trigger an error if it is actually attempted.
4831 The various declarations do not return anything. They store the
4832 various declarations in the parse context.
4834 ###### Parser: grammar
4837 Ocean -> OptNL DeclarationList
4839 ## declare terminals
4847 DeclarationList -> Declaration
4848 | DeclarationList Declaration
4850 Declaration -> ERROR Newlines ${
4851 tok_err(c, // UNTESTED
4852 "error: unhandled parse error", &$1);
4858 ## top level grammar
4862 ### The `const` section
4864 As well as being defined in with the code that uses them, constants can
4865 be declared at the top level. These have full-file scope, so they are
4866 always `InScope`, even before(!) they have been declared. The value of
4867 a top level constant can be given as an expression, and this is
4868 evaluated after parsing and before execution.
4870 A function call can be used to evaluate a constant, but it will not have
4871 access to any program state, once such statement becomes meaningful.
4872 e.g. arguments and filesystem will not be visible.
4874 Constants are defined in a section that starts with the reserved word
4875 `const` and then has a block with a list of assignment statements.
4876 For syntactic consistency, these must use the double-colon syntax to
4877 make it clear that they are constants. Type can also be given: if
4878 not, the type will be determined during analysis, as with other
4881 ###### parse context
4882 struct binode *constlist;
4884 ###### top level grammar
4888 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
4889 | const { SimpleConstList } Newlines
4890 | const IN OptNL ConstList OUT Newlines
4891 | const SimpleConstList Newlines
4893 ConstList -> ConstList SimpleConstLine
4896 SimpleConstList -> SimpleConstList ; Const
4900 SimpleConstLine -> SimpleConstList Newlines
4901 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
4904 CType -> Type ${ $0 = $<1; }$
4908 Const -> IDENTIFIER :: CType = Expression ${ {
4910 struct binode *bl, *bv;
4911 struct var *var = new_pos(var, $ID);
4913 v = var_decl(c, $ID.txt);
4915 v->where_decl = var;
4921 v = var_ref(c, $1.txt);
4922 tok_err(c, "error: name already declared", &$1);
4923 type_err(c, "info: this is where '%v' was first declared",
4924 v->where_decl, NULL, 0, NULL);
4935 bl->left = c->constlist;
4940 ###### core functions
4941 static void resolve_consts(struct parse_context *c)
4944 c->constlist = reorder_bilist(c->constlist);
4945 for (b = cast(binode, c->constlist); b;
4946 b = cast(binode, b->right)) {
4948 struct binode *vb = cast(binode, b->left);
4949 struct var *v = cast(var, vb->left);
4952 propagate_types(vb->right, c, &perr,
4954 } while (perr & Eretry);
4958 struct value res = interp_exec(
4959 c, vb->right, &v->var->type);
4960 global_alloc(c, v->var->type, v->var, &res);
4965 ###### print const decls
4970 for (b = cast(binode, context.constlist); b;
4971 b = cast(binode, b->right)) {
4972 struct binode *vb = cast(binode, b->left);
4973 struct var *vr = cast(var, vb->left);
4974 struct variable *v = vr->var;
4980 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
4981 type_print(v->type, stdout);
4983 print_exec(vb->right, -1, 0);
4988 ###### free const decls
4989 free_binode(context.constlist);
4991 ### Function declarations
4993 The code in an Ocean program is all stored in function declarations.
4994 One of the functions must be named `main` and it must accept an array of
4995 strings as a parameter - the command line arguments.
4997 As this is the top level, several things are handled a bit differently.
4998 The function is not interpreted by `interp_exec` as that isn't passed
4999 the argument list which the program requires. Similarly type analysis
5000 is a bit more interesting at this level.
5002 ###### ast functions
5004 static struct type *handle_results(struct parse_context *c,
5005 struct binode *results)
5007 /* Create a 'struct' type from the results list, which
5008 * is a list for 'struct var'
5010 struct type *t = add_anon_type(c, &structure_prototype,
5011 " function result");
5015 for (b = results; b; b = cast(binode, b->right))
5017 t->structure.nfields = cnt;
5018 t->structure.fields = calloc(cnt, sizeof(struct field));
5020 for (b = results; b; b = cast(binode, b->right)) {
5021 struct var *v = cast(var, b->left);
5022 struct field *f = &t->structure.fields[cnt++];
5023 int a = v->var->type->align;
5024 f->name = v->var->name->name;
5025 f->type = v->var->type;
5027 f->offset = t->size;
5028 v->var->frame_pos = f->offset;
5029 t->size += ((f->type->size - 1) | (a-1)) + 1;
5032 variable_unlink_exec(v->var);
5034 free_binode(results);
5038 static struct variable *declare_function(struct parse_context *c,
5039 struct variable *name,
5040 struct binode *args,
5042 struct binode *results,
5046 struct value fn = {.function = code};
5048 var_block_close(c, CloseFunction, code);
5049 t = add_anon_type(c, &function_prototype,
5050 "func %.*s", name->name->name.len,
5051 name->name->name.txt);
5053 t->function.params = reorder_bilist(args);
5055 ret = handle_results(c, reorder_bilist(results));
5056 t->function.inline_result = 1;
5057 t->function.local_size = ret->size;
5059 t->function.return_type = ret;
5060 global_alloc(c, t, name, &fn);
5061 name->type->function.scope = c->out_scope;
5066 var_block_close(c, CloseFunction, NULL);
5068 c->out_scope = NULL;
5072 ###### declare terminals
5075 ###### top level grammar
5078 DeclareFunction -> func FuncName ( OpenScope ArgsLine ) Block Newlines ${
5079 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5081 | func FuncName IN OpenScope Args OUT OptNL do Block Newlines ${
5082 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5084 | func FuncName NEWLINE OpenScope OptNL do Block Newlines ${
5085 $0 = declare_function(c, $<FN, NULL, Tnone, NULL, $<Bl);
5087 | func FuncName ( OpenScope ArgsLine ) : Type Block Newlines ${
5088 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5090 | func FuncName ( OpenScope ArgsLine ) : ( ArgsLine ) Block Newlines ${
5091 $0 = declare_function(c, $<FN, $<AL, NULL, $<AL2, $<Bl);
5093 | func FuncName IN OpenScope Args OUT OptNL return Type Newlines do Block Newlines ${
5094 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5096 | func FuncName NEWLINE OpenScope return Type Newlines do Block Newlines ${
5097 $0 = declare_function(c, $<FN, NULL, $<Ty, NULL, $<Bl);
5099 | func FuncName IN OpenScope Args OUT OptNL return IN Args OUT OptNL do Block Newlines ${
5100 $0 = declare_function(c, $<FN, $<Ar, NULL, $<Ar2, $<Bl);
5102 | func FuncName NEWLINE OpenScope return IN Args OUT OptNL do Block Newlines ${
5103 $0 = declare_function(c, $<FN, NULL, NULL, $<Ar, $<Bl);
5106 ###### print func decls
5111 while (target != 0) {
5113 for (v = context.in_scope; v; v=v->in_scope)
5114 if (v->depth == 0 && v->type && v->type->check_args) {
5123 struct value *val = var_value(&context, v);
5124 printf("func %.*s", v->name->name.len, v->name->name.txt);
5125 v->type->print_type_decl(v->type, stdout);
5127 print_exec(val->function, 0, brackets);
5129 print_value(v->type, val, stdout);
5130 printf("/* frame size %d */\n", v->type->function.local_size);
5136 ###### core functions
5138 static int analyse_funcs(struct parse_context *c)
5142 for (v = c->in_scope; v; v = v->in_scope) {
5146 if (v->depth != 0 || !v->type || !v->type->check_args)
5148 ret = v->type->function.inline_result ?
5149 Tnone : v->type->function.return_type;
5150 val = var_value(c, v);
5153 propagate_types(val->function, c, &perr, ret, 0);
5154 } while (!(perr & Efail) && (perr & Eretry));
5155 if (!(perr & Efail))
5156 /* Make sure everything is still consistent */
5157 propagate_types(val->function, c, &perr, ret, 0);
5160 if (!v->type->function.inline_result &&
5161 !v->type->function.return_type->dup) {
5162 type_err(c, "error: function cannot return value of type %1",
5163 v->where_decl, v->type->function.return_type, 0, NULL);
5166 scope_finalize(c, v->type);
5171 static int analyse_main(struct type *type, struct parse_context *c)
5173 struct binode *bp = type->function.params;
5177 struct type *argv_type;
5179 argv_type = add_anon_type(c, &array_prototype, "argv");
5180 argv_type->array.member = Tstr;
5181 argv_type->array.unspec = 1;
5183 for (b = bp; b; b = cast(binode, b->right)) {
5187 propagate_types(b->left, c, &perr, argv_type, 0);
5189 default: /* invalid */ // NOTEST
5190 propagate_types(b->left, c, &perr, Tnone, 0); // NOTEST
5196 return !c->parse_error;
5199 static void interp_main(struct parse_context *c, int argc, char **argv)
5201 struct value *progp = NULL;
5202 struct text main_name = { "main", 4 };
5203 struct variable *mainv;
5209 mainv = var_ref(c, main_name);
5211 progp = var_value(c, mainv);
5212 if (!progp || !progp->function) {
5213 fprintf(stderr, "oceani: no main function found.\n");
5217 if (!analyse_main(mainv->type, c)) {
5218 fprintf(stderr, "oceani: main has wrong type.\n");
5222 al = mainv->type->function.params;
5224 c->local_size = mainv->type->function.local_size;
5225 c->local = calloc(1, c->local_size);
5227 struct var *v = cast(var, al->left);
5228 struct value *vl = var_value(c, v->var);
5238 mpq_set_ui(argcq, argc, 1);
5239 memcpy(var_value(c, t->array.vsize), &argcq, sizeof(argcq));
5240 t->prepare_type(c, t, 0);
5241 array_init(v->var->type, vl);
5242 for (i = 0; i < argc; i++) {
5243 struct value *vl2 = vl->array + i * v->var->type->array.member->size;
5245 arg.str.txt = argv[i];
5246 arg.str.len = strlen(argv[i]);
5247 free_value(Tstr, vl2);
5248 dup_value(Tstr, &arg, vl2);
5252 al = cast(binode, al->right);
5254 v = interp_exec(c, progp->function, &vtype);
5255 free_value(vtype, &v);
5260 ###### ast functions
5261 void free_variable(struct variable *v)
5265 ## And now to test it out.
5267 Having a language requires having a "hello world" program. I'll
5268 provide a little more than that: a program that prints "Hello world"
5269 finds the GCD of two numbers, prints the first few elements of
5270 Fibonacci, performs a binary search for a number, and a few other
5271 things which will likely grow as the languages grows.
5273 ###### File: oceani.mk
5276 @echo "===== DEMO ====="
5277 ./oceani --section "demo: hello" oceani.mdc 55 33
5283 four ::= 2 + 2 ; five ::= 10/2
5284 const pie ::= "I like Pie";
5285 cake ::= "The cake is"
5293 func main(argv:[argc::]string)
5294 print "Hello World, what lovely oceans you have!"
5295 print "Are there", five, "?"
5296 print pi, pie, "but", cake
5298 A := $argv[1]; B := $argv[2]
5300 /* When a variable is defined in both branches of an 'if',
5301 * and used afterwards, the variables are merged.
5307 print "Is", A, "bigger than", B,"? ", bigger
5308 /* If a variable is not used after the 'if', no
5309 * merge happens, so types can be different
5312 double:string = "yes"
5313 print A, "is more than twice", B, "?", double
5316 print "double", B, "is", double
5321 if a > 0 and then b > 0:
5327 print "GCD of", A, "and", B,"is", a
5329 print a, "is not positive, cannot calculate GCD"
5331 print b, "is not positive, cannot calculate GCD"
5336 print "Fibonacci:", f1,f2,
5337 then togo = togo - 1
5345 /* Binary search... */
5350 mid := (lo + hi) / 2
5363 print "Yay, I found", target
5365 print "Closest I found was", lo
5370 // "middle square" PRNG. Not particularly good, but one my
5371 // Dad taught me - the first one I ever heard of.
5372 for i:=1; then i = i + 1; while i < size:
5373 n := list[i-1] * list[i-1]
5374 list[i] = (n / 100) % 10 000
5376 print "Before sort:",
5377 for i:=0; then i = i + 1; while i < size:
5381 for i := 1; then i=i+1; while i < size:
5382 for j:=i-1; then j=j-1; while j >= 0:
5383 if list[j] > list[j+1]:
5387 print " After sort:",
5388 for i:=0; then i = i + 1; while i < size:
5392 if 1 == 2 then print "yes"; else print "no"
5396 bob.alive = (bob.name == "Hello")
5397 print "bob", "is" if bob.alive else "isn't", "alive"